Point cloud data encoding device, point cloud data encoding method, point cloud data decoding device, and point cloud data decoding method

Abstract

A method according to embodiments may comprise the steps of: decoding geometry data of point cloud data within a bitstream; and decoding attribute data of the point cloud data. A method according to embodiments may comprise the steps of: encoding geometry data of point cloud data; and encoding attribute data of the point cloud data.

Description

Point cloud data encoding device, point cloud data encoding method, point cloud data decoding device, and point cloud data decoding method

[0001] Embodiments relate to a method and apparatus for processing point cloud content.

[0002] Point cloud content is represented as a point cloud, a collection of points within a coordinate system representing three-dimensional space. Point cloud content can represent three-dimensional media and is used to provide various services such as VR (Virtual Reality), AR (Augmented Reality), MR (Mixed Reality), and autonomous driving services. However, representing point cloud content requires tens to hundreds of thousands of point data. Therefore, a method for efficiently processing massive amounts of point data is required.

[0003] Embodiments provide devices and methods for efficiently processing point cloud data. Embodiments provide methods and devices for processing point cloud data to address latency and encoding / decoding complexity.

[0004] However, the scope of the embodiments is not limited to the technical tasks described above, and the scope of the embodiments may be expanded to other technical tasks that can be inferred by a person skilled in the art based on the entire described content.

[0005] A method according to embodiments may include a step of decoding geometry data of point cloud data in a bitstream; and a step of decoding attribute data of the point cloud data. A method according to embodiments may include a step of encoding geometry data of the point cloud data; and a step of encoding attribute data of the point cloud data.

[0006] The device and method according to the embodiments can process point cloud data with high efficiency.

[0007] The device and method according to the embodiments can provide a high quality point cloud service.

[0008] The device and method according to the embodiments can provide point cloud content for providing general services such as VR services and autonomous driving services.

[0009] The drawings are included to further understand the embodiments, and the drawings illustrate the embodiments together with the description related to the embodiments. For a better understanding of the various embodiments described below, reference should be made to the following description of the embodiments in conjunction with the following drawings, in which like reference numerals correspond to corresponding parts throughout the drawings.

[0010] Figure 1 illustrates an example of a point cloud content provision system according to embodiments.

[0011] FIG. 2 is a block diagram illustrating a point cloud content provision operation according to embodiments.

[0012] FIG. 3 illustrates an example of a point cloud encoder according to embodiments.

[0013] Figure 4 shows examples of octree and occupancy codes according to embodiments.

[0014] Figure 5 shows an example of a point configuration by LOD according to embodiments.

[0015] Figure 6 shows an example of a point configuration by LOD according to embodiments.

[0016] Fig. 7 illustrates an example of a point cloud decoder according to embodiments.

[0017] Figure 8 is an example of a transmission device according to embodiments.

[0018] Fig. 9 is an example of a receiving device according to embodiments.

[0019] Fig. 10 shows an example of a structure that can be linked with a point cloud data transmission / reception method / device according to embodiments.

[0020] Figure 11 is an example of point cloud segmentation according to embodiments.

[0021] Figure 12 shows an example in which the original signal and the reconstructed signal according to the embodiments are included in one cube.

[0022] Figure 13 shows an example of two different point clouds of the same scale, with two spheres having radii of 16 and 2 added for comparison according to the embodiments.

[0023] Figure 14 shows examples of Z-order curves (Lebesgue curve 1904, Morton space-filling curve 1966) according to embodiments.

[0024] Figure 15 shows an example of a 2D projection of a block defined by 8 corner voxels and 27 corner + middle corner + center voxels according to embodiments.

[0025] Figure 16 shows an example of retrieving truncated_block_code from the Morton address of a voxel according to embodiments.

[0026] Figure 17 shows information about a Molton code for neighbor search according to embodiments.

[0027] Figure 18 shows the width distribution of filter taps according to distance (for 2D and 3D cases) according to embodiments.

[0028] Fig. 19 shows the filter shape of HM.7 (reference s / w of HEVC) according to embodiments.

[0029] Figure 20 shows the width distribution of filter taps according to the direction according to the embodiments (3D case).

[0030] Figure 21 shows an example of a 4-tap symmetric ALF design using 8 octants according to embodiments.

[0031] Figure 22 shows a flowchart of resizing and rounding according to embodiments.

[0032] Figure 23 shows a flowchart of calculating ALF tap values ​​according to embodiments.

[0033] Figure 24 shows the dynamic range for calculating a four-tap filter according to embodiments.

[0034] Figure 25 illustrates an RDO-based application of ALF for various QPs according to embodiments.

[0035] Figure 26 shows examples of the most interesting parts within the images according to embodiments.

[0036] Figure 27 illustrates the attribute parameter set data unit syntax in a bitstream according to embodiments.

[0037] Figure 28 illustrates the attribute data unit syntax in a bitstream according to embodiments.

[0038] Figure 29 illustrates the attribute data unit header syntax in a bitstream according to embodiments.

[0039] Fig. 30 shows a flowchart of ALF coefficient coding according to embodiments.

[0040] Fig. 31 shows a flowchart of ALF coefficient coding for four-tap filters encodeCoeff, encodeCoeffLast, and encodeCoeffzeros according to embodiments.

[0041] Figure 32 shows a flowchart of an ALF coefficient decoding process according to embodiments.

[0042] Figures 33a and 33b illustrate attribute data unit ALF coefficient syntax according to embodiments.

[0043] Figure 34 illustrates the attribute ALF coefficient tuple syntax according to embodiments.

[0044] Fig. 35 shows an encoding device according to embodiments.

[0045] Fig. 36 shows a decryption device according to embodiments.

[0046] Figure 37 shows an encoding method according to embodiments.

[0047] Figure 38 shows a decryption method according to embodiments.

[0048] Figure 39 shows an example in which Egyptian_mask_vox12 and cognataQC1 have the same scale according to embodiments.

[0049] Figure 40 shows an example in which Egyptian_mask_vox12 and cognataQC1 have the same scale.

[0050] Figure 41 shows examples of unfilled blocks and blocks after translation (the number of unprocessed voxels after translation reduces the number of unfilled blocks) according to embodiments.

[0051] Figure 42 shows examples of displayed original blocks and unfilled blocks after translation according to embodiments.

[0052] Preferred embodiments of the embodiments are described in detail, examples of which are illustrated in the accompanying drawings. The following detailed description, with reference to the accompanying drawings, is intended to illustrate preferred embodiments of the embodiments, rather than merely show embodiments that can be implemented according to the embodiments. The following detailed description includes details to provide a thorough understanding of the embodiments. However, it will be apparent to those skilled in the art that the embodiments may be practiced without these details.

[0053] While most of the terms used in the examples are commonly used in the field, some terms were arbitrarily selected by the applicant, and their meanings are described in detail in the following descriptions as needed. Therefore, the examples should be understood based on the intended meaning of the terms, not simply their names or meanings.

[0054] Figure 1 illustrates an example of a point cloud content provision system according to embodiments.

[0055] The point cloud content provision system illustrated in FIG. 1 may include a transmission device (10000) and a reception device (10004). The transmission device (10000) and the reception device (10004) are capable of wired and wireless communication to transmit and receive point cloud data.

[0056] A transmission device (10000) according to embodiments can secure, process, and transmit a point cloud video (or point cloud content). According to embodiments, the transmission device (10000) can include a fixed station, a base transceiver system (BTS), a network, an Artificial Intelligence (AI) device and / or system, a robot, an AR / VR / XR device and / or a server, etc. In addition, according to embodiments, the transmission device (10000) can include a device that performs communication with a base station and / or other wireless devices using a wireless access technology (e.g., 5G NR (New RAT), LTE (Long Term Evolution)), a robot, a vehicle, an AR / VR / XR device, a portable device, a home appliance, an IoT (Internet of Things) device, an AI device / server, etc.

[0057] A transmission device (10000) according to embodiments includes a point cloud video acquisition unit (Point Cloud Video Acquisition, 10001), a point cloud video encoder (Point Cloud Video Encoder, 10002), and / or a transmitter (or communication module, 10003).

[0058] A point cloud video acquisition unit (10001) according to embodiments acquires a point cloud video through a processing process such as capture, synthesis, or generation. The point cloud video is point cloud content expressed as a point cloud, which is a collection of points located in a three-dimensional space, and may be referred to as point cloud video data, etc. The point cloud video according to embodiments may include one or more frames. One frame represents a still image / picture. Therefore, the point cloud video may include a point cloud image / frame / picture, and may be referred to as any one of a point cloud image, a frame, and a picture.

[0059] A point cloud video encoder (10002) according to embodiments encodes acquired point cloud video data. The point cloud video encoder (10002) may encode point cloud video data based on point cloud compression coding. The point cloud compression coding according to embodiments may include G-PCC (Geometry-based Point Cloud Compression) coding and / or V-PCC (Video-based Point Cloud Compression) coding or next-generation coding. In addition, the point cloud compression coding according to embodiments is not limited to the above-described embodiment. The point cloud video encoder (10002) may output a bitstream including encoded point cloud video data. The bitstream may include not only encoded point cloud video data but also signaling information related to encoding of the point cloud video data.

[0060] A transmitter (10003) according to embodiments transmits a bitstream including encoded point cloud video data. The bitstream according to embodiments is encapsulated into a file or segment (e.g., streaming segment) and transmitted through various networks such as a broadcast network and / or a broadband network. Although not shown in the drawing, the transmission device (10000) may include an encapsulation unit (or an encapsulation module) that performs an encapsulation operation. In addition, the encapsulation unit may be included in the transmitter (10003) according to embodiments. According to embodiments, the file or segment may be transmitted to a receiving device (10004) through a network or may be stored in a digital storage medium (e.g., USB, SD, CD, DVD, Blu-ray, HDD, SSD, etc.). The transmitter (10003) according to embodiments may communicate with the receiving device (10004) (or receiver (10005)) via a network such as 4G, 5G, or 6G via wired / wireless communication. Additionally, the transmitter (10003) can perform data processing operations required according to a network system (e.g., a communication network system such as 4G, 5G, or 6G). Additionally, the transmission device (10000) can transmit encapsulated data in an on-demand manner.

[0061] A receiving device (10004) according to embodiments includes a receiver (Receiver) 10005, a point cloud video decoder (Point Cloud Decoder) 10006, and / or a renderer (Renderer) 10007. According to embodiments, the receiving device (10004) may include a device, robot, vehicle, AR / VR / XR device, mobile device, home appliance, IoT (Internet of Things) device, AI device / server, etc. that performs communication with a base station and / or other wireless devices using a wireless access technology (e.g., 5G NR (New RAT), LTE (Long Term Evolution)).

[0062] A receiver (10005) according to embodiments receives a bitstream containing point cloud video data or a file / segment in which the bitstream is encapsulated, from a network or a storage medium. The receiver (10005) may perform data processing operations required according to a network system (e.g., a communication network system such as 4G, 5G, or 6G). The receiver (10005) according to embodiments may decapsulate the received file / segment and output a bitstream. In addition, the receiver (10005) according to embodiments may include a decapsulation unit (or decapsulation module) for performing the decapsulation operation. In addition, the decapsulation unit may be implemented as a separate element (or component) from the receiver (10005).

[0063] A point cloud video decoder (10006) decodes a bitstream containing point cloud video data. The point cloud video decoder (10006) can decode the point cloud video data according to how it is encoded (e.g., the reverse process of the operation of the point cloud video encoder (10002)). Accordingly, the point cloud video decoder (10006) can decode the point cloud video data by performing point cloud decompression coding, which is the reverse process of point cloud compression. The point cloud decompression coding includes G-PCC coding.

[0064] The renderer (10007) renders decoded point cloud video data. The renderer (10007) can output point cloud content by rendering not only point cloud video data but also audio data. According to embodiments, the renderer (10007) may include a display for displaying the point cloud content. According to embodiments, the display may not be included in the renderer (10007) but may be implemented as a separate device or component.

[0065] The dotted arrows in the drawing indicate the transmission path of feedback information acquired from the receiving device (10004). The feedback information is information for reflecting the interaction with the user consuming the point cloud content, and includes information about the user (e.g., head orientation information, viewport information, etc.). In particular, when the point cloud content is content for a service requiring interaction with the user (e.g., autonomous driving service, etc.), the feedback information may be transmitted to the content transmitter (e.g., the transmitting device (10000)) and / or the service provider. Depending on the embodiments, the feedback information may be used by the receiving device (10004) as well as the transmitting device (10000), or may not be provided.

[0066] Head orientation information according to embodiments is information about the position, direction, angle, movement, etc. of the user's head. The receiving device (10004) according to embodiments can calculate viewport information based on the head orientation information. The viewport information is information about the area of ​​the point cloud video that the user is looking at. The viewpoint is the point where the user is looking at the point cloud video, and may mean the exact center point of the viewport area. In other words, the viewport is an area centered on the viewpoint, and the size, shape, etc. of the area can be determined by the FOV (Field Of View). Therefore, the receiving device (10004) can extract viewport information based on the vertical or horizontal FOV supported by the device in addition to the head orientation information. In addition, the receiving device (10004) performs gaze analysis, etc. to check the user's point cloud consumption method, the area of ​​the point cloud video that the user is looking at, the gaze time, etc. According to embodiments, the receiving device (10004) may transmit feedback information including gaze analysis results to the transmitting device (10000). The feedback information according to embodiments may be acquired during a rendering and / or display process. The feedback information according to embodiments may be acquired by one or more sensors included in the receiving device (10004). Additionally, according to embodiments, the feedback information may be acquired by a renderer (10007) or a separate external element (or device, component, etc.). The dotted line in Fig. 1 represents a transmission process of feedback information acquired by the renderer (10007). The point cloud content providing system may process (encode / decode) point cloud data based on the feedback information. Therefore, the point cloud video data decoder (10006) may perform a decoding operation based on the feedback information.Additionally, the receiving device (10004) can transmit feedback information to the transmitting device (10000). The transmitting device (10000) (or point cloud video data encoder (10002)) can perform an encoding operation based on the feedback information. Therefore, the point cloud content providing system can efficiently process necessary data (e.g., point cloud data corresponding to the user's head position) based on the feedback information without processing (encoding / decoding) all point cloud data, and provide point cloud content to the user.

[0067] According to embodiments, the transmitting device (10000) may be referred to as an encoder, a transmitting device, a transmitter, etc., and the receiving device (10004) may be referred to as a decoder, a receiving device, a receiver, etc.

[0068] Point cloud data processed (processed through a series of processes of acquisition / encoding / transmission / decoding / rendering) in the point cloud content providing system of FIG. 1 according to embodiments may be referred to as point cloud content data or point cloud video data. According to embodiments, point cloud content data may be used as a concept including metadata or signaling information related to point cloud data.

[0069] The elements of the point cloud content provision system illustrated in FIG. 1 may be implemented by hardware, software, a processor, and / or a combination thereof.

[0070] FIG. 2 is a block diagram illustrating a point cloud content provision operation according to embodiments.

[0071] The block diagram of Fig. 2 illustrates the operation of the point cloud content provision system described in Fig. 1. As described above, the point cloud content provision system can process point cloud data based on point cloud compression coding (e.g., G-PCC).

[0072] A point cloud content providing system according to embodiments (e.g., a point cloud transmission device (10000) or a point cloud video acquisition unit (10001)) can acquire a point cloud video (20000). The point cloud video is expressed as a point cloud belonging to a coordinate system representing a three-dimensional space. The point cloud video according to embodiments can include a Ply (Polygon File format or the Stanford Triangle format) file. If the point cloud video has one or more frames, the acquired point cloud video can include one or more Ply files. The Ply file includes point cloud data such as the geometry and / or attributes of points. The geometry includes the positions of points. The position of each point can be expressed as parameters (e.g., values ​​of each of the X-axis, Y-axis, and Z-axis) representing a three-dimensional coordinate system (e.g., a coordinate system composed of XYZ axes). Attributes include attributes of points (e.g., texture information of each point, color (YCbCr or RGB), reflectance (r), transparency, etc.). A point has one or more attributes (or properties). For example, a point may have one attribute, color, or two attributes, color and reflectance. According to embodiments, geometry may be referred to as positions, geometry information, geometry data, etc., and attributes may be referred to as attributes, attribute information, attribute data, etc.Additionally, a point cloud content provision system (e.g., a point cloud transmission device (10000) or a point cloud video acquisition unit (10001)) can obtain point cloud data from information related to the acquisition process of a point cloud video (e.g., depth information, color information, etc.).

[0073] A point cloud content providing system according to embodiments (e.g., a transmission device (10000) or a point cloud video encoder (10002)) can encode point cloud data (20001). The point cloud content providing system can encode point cloud data based on point cloud compression coding. As described above, point cloud data can include geometry and attributes of points. Therefore, the point cloud content providing system can perform geometry encoding to encode geometry and output a geometry bitstream. The point cloud content providing system can perform attribute encoding to encode attributes and output an attribute bitstream. According to embodiments, the point cloud content providing system can perform attribute encoding based on geometry encoding. The geometry bitstream and the attribute bitstream according to embodiments can be multiplexed and output as a single bitstream. A bitstream according to embodiments may further include signaling information related to geometry encoding and attribute encoding.

[0074] A point cloud content providing system according to embodiments (e.g., a transmission device (10000) or a transmitter (10003)) can transmit encoded point cloud data (20002). As described in FIG. 1, the encoded point cloud data can be expressed as a geometry bitstream and an attribute bitstream. In addition, the encoded point cloud data can be transmitted in the form of a bitstream together with signaling information related to encoding of the point cloud data (e.g., signaling information related to geometry encoding and attribute encoding). In addition, the point cloud content providing system can encapsulate a bitstream that transmits the encoded point cloud data and transmit it in the form of a file or segment.

[0075] A point cloud content providing system according to embodiments (e.g., a receiving device (10004) or a receiver (10005)) can receive a bitstream including encoded point cloud data. In addition, the point cloud content providing system (e.g., a receiving device (10004) or a receiver (10005)) can demultiplex the bitstream.

[0076] A point cloud content providing system (e.g., a receiving device (10004) or a point cloud video decoder (10005)) can decode encoded point cloud data (e.g., a geometry bitstream, an attribute bitstream) transmitted as a bitstream. The point cloud content providing system (e.g., a receiving device (10004) or a point cloud video decoder (10005)) can decode the point cloud video data based on signaling information related to encoding of the point cloud video data included in the bitstream. The point cloud content providing system (e.g., a receiving device (10004) or a point cloud video decoder (10005)) can decode the geometry bitstream to restore positions (geometry) of points. The point cloud content providing system can decode the attribute bitstream based on the restored geometry to restore attributes of points. A point cloud content provision system (e.g., a receiving device (10004) or a point cloud video decoder (10005)) can reconstruct a point cloud video based on positions and decoded attributes according to the reconstructed geometry.

[0077] A point cloud content providing system (e.g., a receiving device (10004) or a renderer (10007)) according to embodiments can render decoded point cloud data (20004). The point cloud content providing system (e.g., a receiving device (10004) or a renderer (10007)) can render the decoded geometry and attributes through a decoding process according to various rendering methods. Points of the point cloud content may be rendered as vertices having a certain thickness, cubes having a certain minimum size centered on the vertex position, or circles centered on the vertex position. All or a portion of the rendered point cloud content is provided to a user through a display (e.g., a VR / AR display, a general display, etc.).

[0078] A point cloud content provision system according to embodiments (e.g., a receiving device (10004)) can obtain feedback information (20005). The point cloud content provision system can encode and / or decode point cloud data based on the feedback information. The feedback information and the operation of the point cloud content provision system according to embodiments are identical to the feedback information and operation described in FIG. 1, and therefore, a detailed description thereof will be omitted.

[0079] FIG. 3 illustrates an example of a point cloud encoder according to embodiments.

[0080] FIG. 3 illustrates an example of a point cloud video encoder (10002) of FIG. 1. The point cloud encoder reconstructs point cloud data (e.g., positions and / or attributes of points) and performs an encoding operation to adjust the quality of point cloud content (e.g., lossless, lossy, near-lossless) depending on network conditions or applications. If the total size of the point cloud content is large (e.g., point cloud content of 60 Gbps at 30 fps), the point cloud content provision system may not be able to stream the content in real time. Therefore, the point cloud content provision system can reconstruct the point cloud content based on the maximum target bitrate in order to provide it according to the network environment, etc.

[0081] As described in FIGS. 1 and 2, the point cloud encoder can perform geometry encoding and attribute encoding. Geometry encoding is performed before attribute encoding.

[0082] The point cloud encoder according to the embodiments includes a coordinate system transformation unit (Transformation Coordinates, 30000), a quantization unit (Quantize and Remove Points (Voxelize), 30001), an octree analysis unit (Analyze Octree, 30002), a surface approximation analysis unit (Analyze Surface Approximation, 30003), an arithmetic encoder (Arithmetic Encode, 30004), a geometry reconstruction unit (Reconstruct Geometry, 30005), a color transformation unit (Transform Colors, 30006), an attribute transformation unit (Transfer Attributes, 30007), a RAHT transformation unit (30008), a LOD generation unit (Generated LOD, 30009), a lifting transformation unit (Lifting) (30010), and a coefficient quantization unit (Quantize Coefficients, 30011) and / or an arithmetic encoder (30012).

[0083] The coordinate system transformation unit (30000), the quantization unit (30001), the octree analysis unit (30002), the surface approximation analysis unit (30003), the arithmetic encoder (30004), and the geometry reconstruction unit (30005) can perform geometry encoding. The geometry encoding according to the embodiments can include octree geometry coding, direct coding, trisoup geometry encoding, and entropy encoding. Direct coding and trisoup geometry encoding are applied selectively or in combination. In addition, the geometry encoding is not limited to the above examples.

[0084] As illustrated in the drawing, a coordinate system conversion unit (30000) according to embodiments receives positions and converts them into coordinates. For example, the positions may be converted into location information of a three-dimensional space (e.g., a three-dimensional space expressed in an XYZ coordinate system, etc.). The location information of the three-dimensional space according to embodiments may be referred to as geometry information.

[0085] A quantization unit (30001) according to embodiments quantizes geometry. For example, the quantization unit (30001) may quantize points based on the minimum position value of all points (e.g., the minimum value on each axis for the X-axis, Y-axis, and Z-axis). The quantization unit (30001) performs a quantization operation of multiplying the difference between the minimum position value and the position value of each point by a preset quantization scale value, and then rounding down or up to find the closest integer value. Accordingly, one or more points may have the same quantized position (or position value). The quantization unit (30001) according to embodiments performs voxelization based on the quantized positions to reconstruct the quantized points. The minimum unit containing two-dimensional image / video information is a pixel, and points of point cloud content (or three-dimensional point cloud video) according to embodiments may be included in one or more voxels. A voxel is a combination of a volume and a pixel, and refers to a three-dimensional cubic space generated when a three-dimensional space is divided into units (unit=1.0) based on axes representing the three-dimensional space (e.g., X-axis, Y-axis, Z-axis). The quantization unit (40001) may match groups of points in the three-dimensional space to voxels. According to embodiments, one voxel may include only one point. According to embodiments, one voxel may include one or more points. In addition, in order to express one voxel as one point, the position of the center of the voxel may be set based on the positions of one or more points included in one voxel. In this case, the attributes of all positions contained in one voxel can be combined and assigned to the voxel.

[0086] The octree analysis unit (30002) according to the embodiments performs octree geometry coding (or octree coding) to represent voxels in an octree structure. The octree structure represents points matched to voxels based on an octree structure.

[0087] The surface approximation analysis unit (30003) according to the embodiments can analyze and approximate an octree. The octree analysis and approximation according to the embodiments is a process of analyzing and voxelizing an area including a large number of points to efficiently provide an octree and voxelization.

[0088] An arithmetic encoder (30004) according to embodiments entropy encodes an octree and / or an approximated octree. For example, the encoding method includes an arithmetic encoding method. The encoding results in a geometry bitstream.

[0089] The color conversion unit (30006), the attribute conversion unit (30007), the RAHT conversion unit (30008), the LOD generation unit (30009), the lifting conversion unit (30010), the coefficient quantization unit (30011) and / or the arithmetic encoder (30012) perform attribute encoding. As described above, one point may have one or more attributes. Attribute encoding according to embodiments is applied equally to the attributes of one point. However, when one attribute (e.g., color) includes one or more elements, independent attribute encoding is applied to each element. Attribute encoding according to embodiments may include color transform coding, attribute transform coding, RAHT (Region Adaptive Hierarchial Transform) coding, Interpolarization-based hierarchical nearest-neighbor prediction-Prediction Transform) coding, and lifting transform (interpolation-based hierarchical nearest-neighbor prediction with an update / lifting step (Lifting Transform)) coding. Depending on the point cloud content, the above-described RAHT coding, prediction transform coding, and lifting transform coding may be selectively used, or a combination of one or more codings may be used. In addition, attribute encoding according to embodiments is not limited to the above-described examples.

[0090] The color conversion unit (30006) according to the embodiments performs color conversion coding to convert color values ​​(or textures) included in attributes. For example, the color conversion unit (30006) may convert the format of color information (e.g., convert from RGB to YCbCr). The operation of the color conversion unit (30006) according to the embodiments may be optionally applied depending on the color values ​​included in the attributes.

[0091] The geometry reconstruction unit (30005) according to the embodiments reconstructs (decompresses) an octree and / or an approximated octree. The geometry reconstruction unit (30005) reconstructs an octree / voxel based on the results of analyzing the distribution of points. The reconstructed octree / voxel may be referred to as a reconstructed geometry (or restored geometry).

[0092] The attribute conversion unit (30007) according to the embodiments performs attribute conversion that converts attributes based on positions for which geometry encoding has not been performed and / or reconstructed geometry. As described above, since the attributes are dependent on the geometry, the attribute conversion unit (30007) can convert the attributes based on the reconstructed geometry information. For example, the attribute conversion unit (30007) can convert the attribute of a point at a position based on the position value of the point included in the voxel. As described above, when the position of the center point of a voxel is set based on the positions of one or more points included in the voxel, the attribute conversion unit (30007) converts the attributes of one or more points. When try-soup geometry encoding is performed, the attribute conversion unit (30007) can convert attributes based on the try-soup geometry encoding.

[0093] The attribute transformation unit (30007) can perform attribute transformation by calculating the average value of the attributes or attribute values ​​(e.g., the color or reflectance of each point) of neighboring points within a specific position / radius from the position (or position value) of the center point of each voxel. The attribute transformation unit (30007) can apply a weight according to the distance from the center point to each point when calculating the average value. Accordingly, each voxel has a position and a calculated attribute (or attribute value).

[0094] The attribute transformation unit (30007) can search for neighboring points within a specific position / radius from the position of the center point of each voxel based on the KD tree or the Moulton code. The KD tree is a binary search tree that supports a data structure that can manage points based on their positions to enable fast nearest neighbor search (NNS). The Moulton code represents the coordinate values ​​(e.g. (x, y, z)) indicating the 3D positions of all points as bit values ​​and is generated by mixing the bits. For example, if the coordinate values ​​indicating the position of a point are (5, 9, 1), the bit values ​​of the coordinate values ​​are (0101, 1001, 0001). If the bit values ​​are mixed in the order of z, y, and x according to the bit index, it is 010001000111. If this value is expressed in decimal, it becomes 1095. That is, the Moulton code value of the point with coordinate values ​​(5, 9, 1) is 1095. The attribute transformation unit (30007) can sort points based on the Moulton code value and perform nearest neighbor search (NNS) through a depth-first traversal process. After the attribute transformation operation, if nearest neighbor search (NNS) is also required in other transformation processes for attribute coding, a KD tree or Moulton code is utilized.

[0095] As shown in the drawing, the converted attributes are input to the RAHT conversion unit (30008) and / or the LOD generation unit (30009).

[0096] The RAHT transform unit (30008) according to the embodiments performs RAHT coding to predict attribute information based on reconstructed geometry information. For example, the RAHT transform unit (30008) can predict attribute information of a node at an upper level of an octree based on attribute information associated with a node at a lower level of the octree.

[0097] The LOD generation unit (30009) according to the embodiments generates a LOD (Level of Detail) to perform predictive transformation coding. The LOD according to the embodiments represents the level of detail of point cloud content. A smaller LOD value indicates lower detail of point cloud content, and a larger LOD value indicates higher detail of point cloud content. Points can be classified according to LOD.

[0098] The lifting transformation unit (30010) according to the embodiments performs lifting transformation coding that transforms attributes of a point cloud based on weights. As described above, lifting transformation coding may be applied selectively.

[0099] The coefficient quantization unit (30011) according to the embodiments quantizes attribute-coded attributes based on coefficients.

[0100] An arithmetic encoder (30012) according to embodiments encodes quantized attributes based on arithmetic coding.

[0101] The elements of the point cloud encoder of FIG. 3 may be implemented by hardware, software, firmware, or a combination thereof, including one or more processors or integrated circuits configured to communicate with one or more memories included in the point cloud providing device, although not shown in the drawing. The one or more processors may perform at least one or more of the operations and / or functions of the elements of the point cloud encoder of FIG. 3 described above. Furthermore, the one or more processors may operate or execute a set of software programs and / or instructions for performing the operations and / or functions of the elements of the point cloud encoder of FIG. 3. The one or more memories according to embodiments may include high-speed random access memory, or may include non-volatile memory (e.g., one or more magnetic disk storage devices, flash memory devices, or other non-volatile solid-state memory devices).

[0102] Figure 4 illustrates examples of octree and occupancy codes according to embodiments.

[0103] As described in FIGS. 1 to 3, the point cloud content provision system (point cloud video encoder (10002)) or point cloud encoder (e.g., octree analysis unit (30002)) performs octree geometry coding (or octree coding) based on an octree structure to efficiently manage the area and / or position of a voxel.

[0104] The top of Fig. 4 shows the octree structure. The three-dimensional space of the point cloud content according to the embodiments is expressed by the axes of the coordinate system (e.g., X-axis, Y-axis, Z-axis). The octree structure has two poles (0,0,0) and (2 d , 2 d , 2 d ) is generated by recursively subdividing the cubical axis-aligned bounding box defined by . 2d can be set to a value that constitutes the smallest bounding box that encloses all points of the point cloud content (or point cloud video). d represents the depth of the octree. The value of d is determined by the following equation. In the equation below, (x int n , y int n , z int n ) represents the positions (or position values) of quantized points.

[0105] d =Ceil(Log2(Max(x_n^int,y_n^int,z_n^int,n=1,…,N)+1))

[0106] As shown in the middle of the upper part of Fig. 4, the entire three-dimensional space can be divided into eight spaces according to the division. Each divided space is expressed as a cube with six faces. As shown in the upper right of Fig. 4, each of the eight spaces is again divided based on the axes of the coordinate system (e.g., X-axis, Y-axis, Z-axis). Therefore, each space is again divided into eight smaller spaces. The divided smaller spaces are also expressed as cubes with six faces. This division method is applied until the leaf nodes of the octree become voxels.

[0107] The bottom of Fig. 4 shows the occupancy code of the octree. The occupancy code of the octree is generated to indicate whether each of the eight partitioned spaces generated by partitioning one space contains at least one point. Therefore, one occupancy code is expressed by eight child nodes. Each child node represents the occupancy of the partitioned space, and each child node has a value of 1 bit. Therefore, the occupancy code is expressed as an 8-bit code. That is, if the space corresponding to the child node contains at least one point, the node has a value of 1. If the space corresponding to the child node does not contain a point (empty), the node has a value of 0. Since the occupancy code illustrated in Fig. 4 is 00100001, it indicates that the spaces corresponding to the third and eighth child nodes among the eight child nodes each contain at least one point. As shown in the drawing, the third child node and the eighth child node each have eight child nodes, and each child node is expressed by an 8-bit occupancy code. The drawing shows that the occupancy code of the third child node is 10000111, and the occupancy code of the eighth child node is 01001111. A point cloud encoder according to embodiments (e.g., an arithmetic encoder (30004)) can entropy encode the occupancy code. In addition, the point cloud encoder can intra / inter code the occupancy code to increase compression efficiency. A receiving device according to embodiments (e.g., a receiving device (10004) or a point cloud video decoder (10006)) reconstructs an octree based on the occupancy code.

[0108] A point cloud encoder according to embodiments (e.g., the point cloud encoder of FIG. 3, or the octree analysis unit (30002)) can perform voxelization and octree coding to store the positions of points. However, points within a 3D space are not always evenly distributed, and thus, there may be specific areas where there are not many points. Therefore, performing voxelization on the entire 3D space is inefficient. For example, if there are few points in a specific area, there is no need to perform voxelization up to that area.

[0109] Therefore, the point cloud encoder according to the embodiments can perform direct coding that directly codes the positions of points included in the specific region (or nodes excluding leaf nodes of the octree) without performing voxelization for the specific region described above. The coordinates of the direct coded points according to the embodiments are referred to as a direct coding mode (DCM). In addition, the point cloud encoder according to the embodiments can perform trisoup geometry encoding that reconstructs the positions of points within the specific region (or node) on a voxel basis based on a surface model. Trisoup geometry encoding is a geometry encoding that expresses the representation of an object as a series of triangle meshes. Therefore, the point cloud decoder can generate a point cloud from the mesh surface. Direct coding and trisoup geometry encoding according to the embodiments can be selectively performed. Additionally, direct coding and tri-subtractive geometry encoding according to embodiments may be performed in combination with octree geometry coding (or octree coding).

[0110] In order to perform direct coding, the option to use direct mode for applying direct coding must be activated, the node to which direct coding is to be applied must not be a leaf node, and there must be points below a threshold within a specific node. In addition, the total number of points subject to direct coding must not exceed a preset threshold. If the above conditions are satisfied, the point cloud encoder (or arithmetic encoder (30004)) according to the embodiments can entropy code the positions (or position values) of the points.

[0111] A point cloud encoder according to embodiments (e.g., surface approximation analysis unit (30003)) can determine a specific level of an octree (when the level is smaller than the depth d of the octree) and, starting from that level, perform tri-subject geometry encoding to reconstruct the positions of points within a node region on a voxel basis using a surface model (tri-subject mode). A point cloud encoder according to embodiments can specify a level to which tri-subject geometry encoding is to be applied. For example, when the specified level is equal to the depth of the octree, the point cloud encoder does not operate in tri-subject mode. That is, a point cloud encoder according to embodiments can operate in tri-subject mode only when the specified level is smaller than the depth value of the octree. A three-dimensional cubic area of ​​nodes at a specified level according to embodiments is called a block. One block may include one or more voxels. A block or a voxel may correspond to a brick. Within each block, geometry is represented by a surface. According to embodiments, a surface may intersect each edge of the block at most once.

[0112] Since one block has 12 edges, there are at least 12 intersections within one block. Each intersection is called a vertex. A vertex existing along an edge is detected if there is at least one occupied voxel adjacent to the edge among all blocks sharing the edge. An occupied voxel according to embodiments means a voxel containing a point. The position of a vertex detected along an edge is the average position along the edge of all voxels adjacent to the edge among all blocks sharing the edge.

[0113] When a vertex is detected, the point cloud encoder according to the embodiments can entropy code the starting point (x, y, z) of the edge, the direction vector (Δx, Δy, Δz) of the edge, and the vertex position value (relative position value within the edge). When the tri-substructure geometry encoding is applied, the point cloud encoder according to the embodiments (e.g., the geometry reconstruction unit (30005)) can perform triangle reconstruction, up-sampling, and voxelization processes to generate restored geometry (reconstructed geometry).

[0114] The vertices located at the edge of a block determine the surface passing through the block. According to the embodiments, the surface is a non-planar polygon. The triangle reconstruction process reconstructs the surface represented by a triangle based on the starting point of the edge, the direction vector of the edge, and the position value of the vertex. The triangle reconstruction process is as follows. ① Calculate the centroid value of each vertex, ② Subtract the centroid value from each vertex value, and ③ Square the values, and then add up all the values ​​to obtain the value.

[0115]

[0116] Find the minimum of the added values, and perform the projection process according to the axis with the minimum value. For example, if the x element is minimum, each vertex is projected to the x-axis based on the center of the block, and then projected onto the (y, z) plane. If the value output when projected onto the (y, z) plane is (ai, bi), the θ value is found through atan2(bi, ai), and the vertices are sorted based on the θ value. The table below shows the combination of vertices to create a triangle depending on the number of vertices. The vertices are sorted in order from 1 to n. The table below shows that two triangles can be formed depending on the combination of vertices for four vertices. The first triangle can be formed by the 1st, 2nd, and 3rd vertices among the sorted vertices, and the second triangle can be formed by the 3rd, 4th, and 1st vertices among the sorted vertices.

[0117] Table 2-1. Triangles formed from vertices ordered 1,x,n

[0118] n triangles

[0119] 3 (1,2,3)

[0120] 4 (1,2,3), (3,4,1)

[0121] 5 (1,2,3), (3,4,5), (5,1,3)

[0122] 6 (1,2,3), (3,4,5), (5,6,1), (1,3,5)

[0123] 7 (1,2,3), (3,4,5), (5,6,7), (7,1,3), (3,5,7)

[0124] 8 (1,2,3), (3,4,5), (5,6,7), (7,8,1), (1,3,5), (5,7,1)

[0125] 9 (1,2,3), (3,4,5), (5,6,7), (7,8,9), (9,1,3), (3,5,7), (7,9,3)

[0126] 10 (1,2,3), (3,4,5), (5,6,7), (7,8,9), (9,10,1), (1,3,5), (5,7,9), (9,1,5)

[0127] 11 (1,2,3), (3,4,5), (5,6,7), (7,8,9), (9,10,11), (11,1,3), (3,5,7), (7,9,11), (11,3,7)

[0128] 12 (1,2,3), (3,4,5), (5,6,7), (7,8,9), (9,10,11), (11,12,1), (1,3,5), (5,7,9), (9,11,1), (1,5,9)

[0129] The upsampling process is performed to voxelize the triangle by adding points in the middle along the edges. Additional points are generated based on the upsampling factor and the width of the block. The additional points are called refined vertices. A point cloud encoder according to embodiments can voxelize the refined vertices. The point cloud encoder can also perform attribute encoding based on the voxelized positions (or position values).

[0130] Figure 5 shows an example of a point configuration by LOD according to embodiments.

[0131] As described in FIGS. 1 to 4, the encoded geometry is reconstructed (decompressed) before attribute encoding is performed. When direct coding is applied, the geometry reconstruction operation may include changing the arrangement of direct-coded points (e.g., placing the direct-coded points at the front of the point cloud data). When trysoup geometry encoding is applied, the geometry reconstruction process includes triangle reconstruction, upsampling, and voxelization. Since attributes depend on the geometry, attribute encoding is performed based on the reconstructed geometry.

[0132] A point cloud encoder (e.g., LOD generation unit (30009)) can reorganize points by LOD. The drawing shows point cloud content corresponding to LOD. The left side of the drawing shows the original point cloud content. The second figure from the left in the drawing shows the distribution of points of the lowest LOD, and the rightmost figure in the drawing shows the distribution of points of the highest LOD. That is, points of the lowest LOD are sparsely distributed, and points of the highest LOD are densely distributed. That is, as LOD increases in the direction of the arrow indicated at the bottom of the drawing, the interval (or distance) between points becomes shorter.

[0133] Figure 6 shows an example of a point configuration by LOD according to embodiments.

[0134] As described in FIGS. 1 to 5, a point cloud content providing system, or a point cloud encoder (e.g., a point cloud video encoder (10002), the point cloud encoder of FIG. 3, or a LOD generation unit (30009)) can generate a LOD. The LOD is generated by reorganizing points into a set of refinement levels according to a set LOD distance value (or a set of Euclidean distances). The LOD generation process is performed not only in the point cloud encoder but also in the point cloud decoder.

[0135] The upper part of Fig. 6 shows examples of points (P0 to P9) of point cloud content distributed in 3D space. The original order in Fig. 6 represents the order of points P0 to P9 before LOD generation. The LOD-based order in Fig. 6 represents the order of points according to LOD generation. The points are rearranged by LOD. Additionally, a higher LOD includes points belonging to a lower LOD. As shown in Fig. 6, LOD0 includes P0, P5, P4, and P2. LOD1 includes points of LOD0 and P1, P6, and P3. LOD2 includes points of LOD0, points of LOD1, and P9, P8, and P7.

[0136] As described in FIG. 3, the point cloud encoder according to the embodiments can selectively or in combination perform predictive transform coding, lifting transform coding, and RAHT transform coding.

[0137] A point cloud encoder according to embodiments can perform predictive transformation coding to generate a predictor for points and set a predicted attribute (or predicted attribute value) for each point. That is, N predictors can be generated for N points. The predictor according to embodiments can calculate a weight (= 1 / distance) value based on the LOD value of each point, indexing information for neighboring points existing within a distance set for each LOD, and distance values ​​to the neighboring points.

[0138] According to the embodiments, the predicted attribute (or attribute value) is set as the average value of the product of the attributes (or attribute values, for example, color, reflectance, etc.) of neighboring points set in the predictor of each point and the weight (or weight value) calculated based on the distance to each neighboring point. The point cloud encoder according to the embodiments (for example, the coefficient quantization unit (30011)) can quantize and inverse quantize the residual values ​​(which may be referred to as residual attributes, residual attribute values, attribute prediction residual values, etc.) obtained by subtracting the predicted attribute (attribute value) from the attribute (attribute value) of each point. The quantization process is as shown in the following table.

[0139] graph. Attribute prediction residuals quantization pseudo code

[0140] int PCCQuantization(int value, int quantStep) {

[0141] if( value >=0) {

[0142] return floor(value / quantStep + 1.0 / 3.0);

[0143] } else {

[0144] return -floor(-value / quantStep + 1.0 / 3.0);

[0145] }

[0146] }

[0147] graph. Attribute prediction residuals inverse quantization pseudo code

[0148] int PCCInverseQuantization(int value, int quantStep) {

[0149] if( quantStep ==0) {

[0150] return value;

[0151] } else {

[0152] return value * quantStep;

[0153] }

[0154] }

[0155] A point cloud encoder according to embodiments (e.g., an arithmetic encoder (30012)) can entropy code the quantized and dequantized residuals as described above when there are neighboring points to the predictor of each point. A point cloud encoder according to embodiments (e.g., an arithmetic encoder (30012)) can entropy code the attributes of the point without performing the above-described process when there are no neighboring points to the predictor of each point.

[0156] A point cloud encoder according to embodiments (e.g., lifting transformation unit (30010)) can perform lifting transformation coding by generating a predictor for each point, setting the LOD calculated in the predictor, registering neighboring points, and setting weights according to the distance to the neighboring points. Lifting transformation coding according to embodiments is similar to the above-described predictive transformation coding, but differs in that weights are cumulatively applied to attribute values. The process of cumulatively applying weights to attribute values ​​according to embodiments is as follows.

[0157] 1) Create an array QW (QuantizationWieght) that stores the weight values ​​of each point. The initial value of all elements in QW is 1.0. Add the value obtained by multiplying the weight of the current point's predictor by the QW value of the predictor index of the neighboring node registered in the predictor.

[0158] 2) Lift prediction process: To calculate the predicted attribute value, the weighted value of the point's attribute value is multiplied and subtracted from the existing attribute value.

[0159] 3) Create temporary arrays called updateweight and update and initialize them to 0.

[0160] 4) For each predictor, the calculated weights are multiplied by the weights stored in the QW corresponding to the predictor index, and the resulting weights are cumulatively added to the update weight array as the index of the neighboring node. The update array accumulates the values ​​obtained by multiplying the calculated weights by the attribute values ​​of the indexes of the neighboring nodes.

[0161] 5) Lift update process: For each predictor, the attribute values ​​in the update array are divided by the weight values ​​in the update weight array of the predictor index, and the existing attribute values ​​are added to the divided value.

[0162] 6) For all predictors, the predicted attribute values ​​are calculated by additionally multiplying the updated attribute values ​​through the lift update process by the weights (stored in QW) updated through the lift prediction process. The point cloud encoder according to the embodiments (e.g., coefficient quantization unit (30011)) quantizes the predicted attribute values. In addition, the point cloud encoder (e.g., arithmetic encoder (30012)) entropy-codes the quantized attribute values.

[0163] A point cloud encoder according to embodiments (e.g., RAHT transform unit (30008)) can perform RAHT transform coding that predicts attributes of upper-level nodes using attributes associated with nodes at lower levels of an octree. RAHT transform coding is an example of attribute intra coding through octree backward scan. A point cloud encoder according to embodiments scans from a voxel to the entire area, and repeats the merging process up to a root node while merging voxels into larger blocks at each step. The merging process according to embodiments is performed only for occupied nodes. The merging process is not performed for empty nodes, and the merging process is performed for the node immediately above the empty node.

[0164] The following equation represents the RAHT transformation matrix. g l x, y, z represents the average attribute value of voxels at level l. g l x, y, z can be calculated from gl+1 2x, y, z and gl+1 2x+1, y, z. The weights of gl 2x, y, z and gl 2x+1, y, z are w1=wl 2x, y, z and w2=wl 2x+1, y, z.

[0165]

[0166] gl-1 x, y, z are low-pass values, which are used in the merging process at the next higher level. hl-1 x, y, z are high-pass coefficients, and the high-pass coefficients at each step are quantized and entropy coded (e.g., encoding of an arithmetic encoder (400012)). The weights are computed as wl-1 x, y, z=wl 2x, y, z+wl 2x+1, y, z. The root node is the last g 1 0, 0, 0 and g 1 0, 0, 1 It is generated as follows:

[0167]

[0168] The gDC values ​​are also quantized and entropy coded, like the high-pass coefficients.

[0169] Fig. 7 illustrates an example of a point cloud decoder according to embodiments.

[0170] The point cloud decoder illustrated in FIG. 7 is an example of a point cloud decoder and can perform a decoding operation, which is the reverse process of the encoding operation of the point cloud encoder described in FIGS. 1 to 6.

[0171] As described in Figure 1, the point cloud decoder can perform geometry decoding and attribute decoding. Geometry decoding is performed before attribute decoding.

[0172] A point cloud decoder according to embodiments includes an arithmetic decoder (7000), an octree synthesizer (7001), a surface approximation synthesizer (7002), a geometry reconstructor (7003), an inverse transform coordinates (7004), an arithmetic decoder (7005), an inverse quantize (7006), a RAHT transform (7007), a LOD generator (7008), an inverse lifting (7009), and / or an inverse transform colors (7010).

[0173] The arithmetic decoder (7000), the octree synthesis unit (7001), the surface oproximation synthesis unit (7002), the geometry reconstruction unit (7003), and the coordinate system inversion unit (7004) can perform geometry decoding. Geometry decoding according to embodiments can include direct coding and trisoup geometry decoding. Direct coding and trisoup geometry decoding are applied selectively. In addition, geometry decoding is not limited to the above examples, and is performed by the reverse process of the geometry encoding described in FIGS. 1 to 6.

[0174] An arithmetic decoder (7000) according to embodiments decodes a received geometry bitstream based on arithmetic coding. The operation of the arithmetic decoder (7000) corresponds to the reverse process of the arithmetic encoder (30004).

[0175] The octree synthesis unit (7001) according to the embodiments can generate an octree by obtaining an occupancy code from a decoded geometry bitstream (or information about the geometry obtained as a result of decoding). A specific description of the occupancy code is as described in FIGS. 1 to 6.

[0176] The surface off-axis synthesis unit (7002) according to the embodiments can synthesize a surface based on the decoded geometry and / or the generated octree when the tri-sub geometry encoding is applied.

[0177] The geometry reconstruction unit (7003) according to the embodiments can regenerate geometry based on the surface and / or decoded geometry. As described in FIGS. 1 to 6, direct coding and try-soup geometry encoding are selectively applied. Therefore, the geometry reconstruction unit (7003) directly retrieves and adds position information of points to which direct coding is applied. In addition, when try-soup geometry encoding is applied, the geometry reconstruction unit (7003) can restore geometry by performing a reconstruction operation of the geometry reconstruction unit (30005), such as triangle reconstruction, up-sampling, and voxelization operations. The specific details are the same as described in FIG. 4 and are therefore omitted. The restored geometry may include a point cloud picture or frame that does not include attributes.

[0178] The coordinate system inverse transformation unit (7004) according to the embodiments can obtain the positions of points by transforming the coordinate system based on the restored geometry.

[0179] The arithmetic decoder (7005), the inverse quantization unit (7006), the RAHT transform unit (7007), the LOD generation unit (7008), the inverse lifting unit (7009), and / or the color inverse transform unit (7010) can perform the attribute decoding described in FIG. 10. The attribute decoding according to embodiments can include RAHT (Region Adaptive Hierarchial Transform) decoding, Interpolaration-based hierarchical nearest-neighbor prediction-Prediction Transform) decoding, and lifting transform (interpolation-based hierarchical nearest-neighbor prediction with an update / lifting step (Lifting Transform)) decoding. The three decodings described above can be used selectively, or a combination of one or more decodings can be used. Additionally, attribute decoding according to embodiments is not limited to the examples described above.

[0180] An arithmetic decoder (7005) according to embodiments decodes an attribute bitstream using arithmetic coding.

[0181] The inverse quantization unit (7006) according to the embodiments inverse quantizes information about the decoded attribute bitstream or the attributes obtained as a result of the decoding and outputs the inverse quantized attributes (or attribute values). The inverse quantization may be selectively applied based on the attribute encoding of the point cloud encoder.

[0182] According to embodiments, the RAHT transform unit (7007), the LOD generator (7008), and / or the inverse lifting unit (7009) may process the reconstructed geometry and the inverse quantized attributes. As described above, the RAHT transform unit (7007), the LOD generator (7008), and / or the inverse lifting unit (7009) may selectively perform a corresponding decoding operation according to the encoding of the point cloud encoder.

[0183] The color inverse transform unit (7010) according to the embodiments performs inverse transform coding to inversely transform the color values ​​(or textures) included in the decoded attributes. The operation of the color inverse transform unit (7010) may be selectively performed based on the operation of the color transform unit (30006) of the point cloud encoder.

[0184] The elements of the point cloud decoder of FIG. 7 may be implemented by hardware, software, firmware, or a combination thereof, including one or more processors or integrated circuits configured to communicate with one or more memories included in a point cloud providing device, although not shown in the drawing. The one or more processors may perform at least one or more of the operations and / or functions of the elements of the point cloud decoder of FIG. 7 described above. Furthermore, the one or more processors may operate or execute a set of software programs and / or instructions for performing the operations and / or functions of the elements of the point cloud decoder of FIG. 7.

[0185] Figure 8 is an example of a transmission device according to embodiments.

[0186] The transmission device illustrated in FIG. 8 is an example of the transmission device (10000) of FIG. 1 (or the point cloud encoder of FIG. 3). The transmission device illustrated in FIG. 8 can perform at least one or more of the same or similar operations and encoding methods as the operations and encoding methods of the point cloud encoder described in FIGS. 1 to 6. A transmission device according to embodiments may include a data input unit (8000), a quantization processing unit (8001), a voxelization processing unit (8002), an octree occupancy code generation unit (8003), a surface model processing unit (8004), an intra / inter coding processing unit (8005), an arithmetic coder (8006), a metadata processing unit (8007), a color conversion processing unit (8008), an attribute conversion processing unit (or a property conversion processing unit) (8009), a prediction / lifting / RAHT conversion processing unit (8010), an arithmetic coder (8011), and / or a transmission processing unit (8012).

[0187] The data input unit (8000) according to the embodiments receives or acquires point cloud data. The data input unit (8000) may perform operations and / or acquisition methods identical or similar to those of the point cloud video acquisition unit (10001) (or the acquisition process (20000) described in FIG. 2).

[0188] The data input unit (8000), quantization processing unit (8001), voxelization processing unit (8002), octree occupancy code generation unit (8003), surface model processing unit (8004), intra / inter coding processing unit (8005), and arithmetic coder (8006) perform geometry encoding. Since the geometry encoding according to the embodiments is the same or similar to the geometry encoding described in FIGS. 1 to 6, a detailed description thereof will be omitted.

[0189] The quantization processing unit (8001) according to the embodiments quantizes geometry (e.g., position values ​​of points or position values). The operation and / or quantization of the quantization processing unit (8001) is identical to or similar to the operation and / or quantization of the quantization unit (30001) described in FIG. 3. The specific description is the same as that described in FIGS. 1 to 6.

[0190] The voxelization processing unit (8002) according to the embodiments voxels the position values ​​of quantized points. The voxelization processing unit (80002) may perform operations and / or processes identical or similar to the operations and / or voxelization processes of the quantization unit (30001) described in FIG. 3. Specific descriptions are identical to those described in FIGS. 1 to 6.

[0191] The octree occupancy code generation unit (8003) according to the embodiments performs octree coding on the positions of voxelized points based on the octree structure. The octree occupancy code generation unit (8003) can generate an occupancy code. The octree occupancy code generation unit (8003) can perform operations and / or methods identical or similar to those of the point cloud encoder (or octree analysis unit (30002)) described in FIGS. 3 and 4. The specific description is the same as that described in FIGS. 1 to 6.

[0192] The surface model processing unit (8004) according to the embodiments can perform tri-subject geometry encoding to reconstruct the positions of points within a specific area (or node) on a voxel basis based on the surface model. The surface model processing unit (8004) can perform operations and / or methods identical or similar to those of the point cloud encoder (e.g., surface approximation analysis unit (30003)) described in FIG. 3. The specific description is the same as that described with reference to FIGS. 1 to 6.

[0193] The intra / inter coding processing unit (8005) according to embodiments can intra / inter code point cloud data. The intra / inter coding processing unit (8005) can perform coding identical to or similar to the intra / inter coding described in FIG. 7. The specific description is identical to that described in FIG. 7. According to embodiments, the intra / inter coding processing unit (8005) can be included in an arithmetic coder (8006).

[0194] An arithmetic coder (8006) according to embodiments entropy encodes an octree and / or an approximated octree of point cloud data. For example, the encoding method includes an arithmetic encoding method. The arithmetic coder (8006) performs operations and / or methods identical or similar to those of the arithmetic encoder (30004).

[0195] The metadata processing unit (8007) according to the embodiments processes metadata regarding point cloud data, such as setting values, and provides the metadata to a necessary processing step, such as geometry encoding and / or attribute encoding. In addition, the metadata processing unit (8007) according to the embodiments may generate and / or process signaling information related to geometry encoding and / or attribute encoding. The signaling information according to the embodiments may be encoded and processed separately from geometry encoding and / or attribute encoding. In addition, the signaling information according to the embodiments may be interleaved.

[0196] The color conversion processing unit (8008), the attribute conversion processing unit (8009), the prediction / lifting / RAHT conversion processing unit (8010), and the arithmetic coder (8011) perform attribute encoding. Since the attribute encoding according to the embodiments is the same as or similar to the attribute encoding described in FIGS. 1 to 6, a detailed description thereof will be omitted.

[0197] The color conversion processing unit (8008) according to the embodiments performs color conversion coding to convert the color values ​​included in the attributes. The color conversion processing unit (8008) can perform color conversion coding based on the reconstructed geometry. The description of the reconstructed geometry is the same as that described with reference to FIGS. 1 to 6. In addition, the color conversion processing unit (8008) performs the same or similar operation and / or method as that of the color conversion unit (30006) described with reference to FIG. 3. A detailed description thereof will be omitted.

[0198] The attribute transformation processing unit (8009) according to embodiments performs attribute transformation to transform attributes based on positions for which geometry encoding has not been performed and / or reconstructed geometry. The attribute transformation processing unit (8009) performs operations and / or methods that are the same as or similar to those of the attribute transformation unit (30007) described in FIG. 3. A detailed description thereof will be omitted. The prediction / lifting / RAHT transformation processing unit (8010) according to embodiments can code transformed attributes by using any one or a combination of RAHT coding, prediction transformation coding, and lifting transformation coding. The prediction / lifting / RAHT transformation processing unit (8010) performs at least one or more of operations that are the same as or similar to those of the RAHT transformation unit (30008), LOD generation unit (30009), and lifting transformation unit (30010) described in FIG. 3. In addition, the description of the prediction transformation coding, lifting transformation coding, and RAHT transformation coding is the same as that described in FIGS. 1 to 6, so a detailed description is omitted.

[0199] An arithmetic coder (8011) according to embodiments can encode coded attributes based on arithmetic coding. The arithmetic coder (8011) performs operations and / or methods identical or similar to those of the arithmetic encoder (300012).

[0200] The transmission processing unit (8012) according to embodiments may transmit each bitstream including encoded geometry and / or encoded attribute, metadata information, or may transmit the encoded geometry and / or encoded attribute, and metadata information as one bitstream. When the encoded geometry and / or encoded attribute, and metadata information according to embodiments are configured as one bitstream, the bitstream may include one or more sub-bitstreams. The bitstream according to embodiments may include signaling information including a Sequence Parameter Set (SPS) for sequence-level signaling, a Geometry Parameter Set (GPS) for signaling geometry information coding, an Attribute Parameter Set (APS) for signaling attribute information coding, and a Tile Parameter Set (TPS) for tile-level signaling, and slice data. The slice data may include information about one or more slices. One slice according to embodiments may include one geometry bitstream (Geom0). 0 ) and one or more attribute bitstreams (Attr0 0 , Attr1 0 ) may be included.

[0201] A slice is a series of syntax elements that represent all or part of a coded point cloud frame.

[0202] A TPS according to embodiments may include information about each tile (e.g., coordinate value information of a bounding box and height / size information, etc.) for one or more tiles. A geometry bitstream may include a header and a payload. The header of a geometry bitstream according to embodiments may include identification information of a parameter set included in GPS (geom_ parameter_set_id), a tile identifier (geom_tile_id), a slice identifier (geom_slice_id), and information about data included in the payload. As described above, a metadata processing unit (8007) according to embodiments may generate and / or process signaling information and transmit it to a transmission processing unit (8012). According to embodiments, elements that perform geometry encoding and elements that perform attribute encoding may share data / information with each other as indicated by a dotted line. The transmission processing unit (8012) according to the embodiments may perform operations and / or transmission methods identical or similar to those of the transmitter (10003). A detailed description thereof is omitted as it is the same as that described in FIGS. 1 and 2.

[0203] Fig. 9 is an example of a receiving device according to embodiments.

[0204] The receiving device illustrated in FIG. 9 is an example of the receiving device (10004) of FIG. 1 (or the point cloud decoder of FIGS. 10 and 11). The receiving device illustrated in FIG. 9 can perform at least one or more of the same or similar operations and decoding methods as the operations and decoding methods of the point cloud decoder described in FIGS. 1 to 11.

[0205] A receiving device according to embodiments may include a receiving unit (9000), a receiving processing unit (9001), an arithmetic decoder (9002), an occupancy code-based octree reconstruction processing unit (9003), a surface model processing unit (triangle reconstruction, up-sampling, voxelization) (9004), an inverse quantization processing unit (9005), a metadata parser (9006), an arithmetic decoder (9007), an inverse quantization processing unit (9008), a prediction / lifting / RAHT inverse transform processing unit (9009), a color inverse transform processing unit (9010), and / or a renderer (9011). Each component of the decoding according to embodiments may perform the reverse process of the component of the encoding according to embodiments.

[0206] The receiving unit (9000) according to the embodiments receives point cloud data. The receiving unit (9000) may perform operations and / or receiving methods identical or similar to those of the receiver (10005) of FIG. 1. A detailed description thereof will be omitted.

[0207] The receiving processing unit (9001) according to the embodiments can obtain a geometry bitstream and / or an attribute bitstream from the received data. The receiving processing unit (9001) can be included in the receiving unit (9000).

[0208] The arithmetic decoder (9002), the occupancy code-based octree reconstruction processing unit (9003), the surface model processing unit (9004), and the inverse quantization processing unit (9005) can perform geometry decoding. Since the geometry decoding according to the embodiments is the same or similar to the geometry decoding described in FIGS. 1 to 10, a detailed description thereof will be omitted.

[0209] An arithmetic decoder (9002) according to embodiments can decode a geometry bitstream based on arithmetic coding. The arithmetic decoder (9002) performs operations and / or coding identical to or similar to those of the arithmetic decoder (7000).

[0210] The occupancy code-based octree reconstruction processing unit (9003) according to embodiments can reconstruct an octree by obtaining an occupancy code from a decoded geometry bitstream (or information about the geometry obtained as a result of decoding). The occupancy code-based octree reconstruction processing unit (9003) performs the same or similar operations and / or methods as those of the octree synthesis unit (7001) and / or the octree generation method. The surface model processing unit (9004) according to embodiments can perform tri-sub geometry decoding and related geometry reconstructing (e.g., triangle reconstruction, up-sampling, voxelization) based on the surface model method when tri-sub geometry encoding is applied. The surface model processing unit (9004) performs the same or similar operations as those of the surface off-ratio synthesis unit (7002) and / or the geometry reconstructing unit (7003).

[0211] The inverse quantization processing unit (9005) according to the embodiments can inverse quantize the decoded geometry.

[0212] The metadata parser (9006) according to the embodiments can parse metadata, such as setting values, contained in the received point cloud data. The metadata parser (9006) can pass the metadata to geometry decoding and / or attribute decoding. A detailed description of the metadata is omitted as it is the same as described in FIG. 8.

[0213] The arithmetic decoder (9007), the inverse quantization processing unit (9008), the prediction / lifting / RAHT inverse transform processing unit (9009), and the color inverse transform processing unit (9010) perform attribute decoding. Since attribute decoding is the same or similar to the attribute decoding described in FIGS. 1 to 10, a detailed description thereof will be omitted.

[0214] An arithmetic decoder (9007) according to embodiments can decode an attribute bitstream using arithmetic coding. The arithmetic decoder (9007) can decode the attribute bitstream based on the reconstructed geometry. The arithmetic decoder (9007) performs operations and / or coding identical or similar to those of the arithmetic decoder (7005).

[0215] The inverse quantization processing unit (9008) according to the embodiments can inverse quantize the decoded attribute bitstream. The inverse quantization processing unit (9008) performs operations and / or methods identical or similar to the operations and / or inverse quantization methods of the inverse quantization unit (7006).

[0216] The prediction / lifting / RAHT inverse transform processing unit (9009) according to embodiments can process reconstructed geometry and inverse quantized attributes. The prediction / lifting / RAHT inverse transform processing unit (9009) performs at least one or more of operations and / or decodings that are identical or similar to the operations and / or decodings of the RAHT transform unit (7007), the LOD generation unit (7008), and / or the inverse lifting unit (7009). The color inverse transform processing unit (9010) according to embodiments performs inverse transform coding for inverse transforming the color value (or texture) included in the decoded attributes. The color inverse transform processing unit (9010) performs operations and / or inverse transform coding that are identical or similar to the operations and / or inverse transform coding of the color inverse transform unit (7010). A renderer (9011) according to embodiments can render point cloud data.

[0217] Fig. 10 shows an example of a structure that can be linked with a point cloud data transmission / reception method / device according to embodiments.

[0218] The structure of FIG. 10 represents a configuration in which at least one of a server (1060), a robot (1010), an autonomous vehicle (1020), an XR device (1030), a smartphone (1040), a home appliance (1050), and / or an HMD (1070) is connected to a cloud network (1010). The robot (1010), the autonomous vehicle (1020), the XR device (1030), the smartphone (1040), or the home appliance (1050) are referred to as devices. In addition, the XR device (1030) may correspond to or be linked with a point cloud data (PCC) device according to embodiments.

[0219] A cloud network (1000) may refer to a network that constitutes part of a cloud computing infrastructure or exists within the cloud computing infrastructure. Here, the cloud network (1000) may be configured using a 3G network, a 4G or LTE (Long Term Evolution) network, or a 5G network.

[0220] The server (1060) is connected to at least one of a robot (1010), an autonomous vehicle (1020), an XR device (1030), a smartphone (1040), a home appliance (1050), and / or an HMD (1070) through a cloud network (1000), and can assist in at least part of the processing of the connected devices (1010 to 1070).

[0221] The HMD (Head-Mount Display) (1070) represents one of the types in which the XR device and / or the PCC device according to the embodiments can be implemented. The HMD type device according to the embodiments includes a communication unit, a control unit, a memory unit, an I / O unit, a sensor unit, and a power supply unit.

[0222] Below, various embodiments of devices (1010 to 1050) to which the above-described technology is applied are described. Here, the devices (1010 to 1050) illustrated in FIG. 10 can be linked / combined with point cloud data transmission / reception devices according to the above-described embodiments.

[0223] <PCC+XR>

[0224] The XR / PCC device (1030) may be implemented as a HMD (Head-Mount Display), a HUD (Head-Up Display) equipped in a vehicle, a television, a mobile phone, a smart phone, a computer, a wearable device, a home appliance, digital signage, a vehicle, a fixed robot, a mobile robot, etc., by applying PCC and / or XR (AR+VR) technology.

[0225] The XR / PCC device (1030) can obtain information about surrounding space or real objects by analyzing 3D point cloud data or image data acquired through various sensors or from external devices to generate location data and attribute data for 3D points, and can render and output an XR object to be output. For example, the XR / PCC device (1030) can output an XR object including additional information about a recognized object in correspondence with the recognized object.

[0226] <PCC+XR+모바일폰>

[0227] The XR / PCC device (1030) can be implemented as a mobile phone (1040) or the like by applying PCC technology.

[0228] The mobile phone (1040) can decode and display point cloud content based on PCC technology.

[0229] <PCC+자율주행+XR>

[0230] Autonomous vehicles (1020) can be implemented as mobile robots, vehicles, unmanned aerial vehicles, etc. by applying PCC technology and XR technology.

[0231] An autonomous vehicle (1020) to which XR / PCC technology is applied may refer to an autonomous vehicle equipped with a means for providing XR images, or an autonomous vehicle that is the subject of control / interaction within an XR image. In particular, an autonomous vehicle (1020) that is the subject of control / interaction within an XR image is distinct from an XR device (1030) and can be linked with each other.

[0232] An autonomous vehicle (1020) equipped with a means for providing XR / PCC images can obtain sensor information from sensors including cameras and output XR / PCC images generated based on the obtained sensor information. For example, the autonomous vehicle (1020) can be equipped with a HUD to output XR / PCC images, thereby providing passengers with XR / PCC objects corresponding to real objects or objects on a screen.

[0233] At this time, when the XR / PCC object is output to the HUD, at least a part of the XR / PCC object may be output so as to overlap with an actual object toward which the passenger's gaze is directed. On the other hand, when the XR / PCC object is output to a display provided inside the autonomous vehicle, at least a part of the XR / PCC object may be output so as to overlap with an object on the screen. For example, the autonomous vehicle (1220) may output XR / PCC objects corresponding to objects such as a lane, another vehicle, a traffic light, a traffic sign, a two-wheeled vehicle, a pedestrian, a building, etc.

[0234] VR (Virtual Reality) technology, AR (Augmented Reality) technology, MR (Mixed Reality) technology and / or PCC (Point Cloud Compression) technology according to the embodiments can be applied to various devices.

[0235] In other words, VR technology is a display technology that provides only CG images of objects or backgrounds in the real world. On the other hand, AR technology refers to a technology that shows a virtually created CG image on top of an image of an actual object. Furthermore, MR technology is similar to the aforementioned AR technology in that it mixes and combines virtual objects in the real world. However, in AR technology, the distinction between real objects and virtual objects created with CG images is clear, and virtual objects are used in a form that complements real objects, whereas in MR technology, virtual objects are considered to have the same characteristics as real objects. A more specific example is the hologram service, which is an application of the aforementioned MR technology.

[0236] However, recently, rather than clearly distinguishing between VR, AR, and MR technologies, they are often referred to as XR (extended reality) technologies. Therefore, embodiments of the present invention are applicable to all VR, AR, MR, and XR technologies. These technologies can be applied to encoding / decoding based on PCC, V-PCC, and G-PCC technologies.

[0237] The PCC method / device according to the embodiments can be applied to a vehicle providing an autonomous driving service.

[0238] Vehicles providing autonomous driving services are connected to PCC devices to enable wired / wireless communication.

[0239] A point cloud data (PCC) transmission and reception device according to embodiments, when connected to a vehicle to enable wired / wireless communication, can receive / process content data related to AR / VR / PCC services that can be provided together with autonomous driving services and transmit the same to the vehicle. In addition, when the point cloud data transmission and reception device is mounted on a vehicle, the point cloud transmission and reception device can receive / process content data related to AR / VR / PCC services and provide the same to a user according to a user input signal input through a user interface device. A vehicle or a user interface device according to embodiments can receive a user input signal. The user input signal according to embodiments can include a signal instructing an autonomous driving service.

[0240] The encoding method and device according to the embodiments may include, for example, Fig. 1 transmitting device (10000), encoder (10002), Fig. 2 obtaining (20000) to transmitting (20002), Fig. 4, Fig. 12 encoder, Fig. 10 each device, Fig. 11 to Fig. 26 ALF (Adaptive Loop Filter) operation, Fig. 27 to Fig. 29, Fig. 33 to Fig. 34 HLS (High Level Syntax) generation, Fig. 30 to Fig. 31 encoding, Fig. 35 encoder, Fig. 37 encoding method, etc.

[0241] The decoding method and device according to the embodiments may include a receiving device (10004) in FIG. 1, a decoder (10006), a transmission (20002) to a rendering (20004) in FIG. 2, a decoder in FIG. 7, a decoder in FIG. 13, each device in FIG. 10, an ALF (Adaptive Loop Filter) operation in FIG. 11 to FIG. 26, an HLS (High Level Syntax) acquisition in FIG. 27 to FIG. 29, and FIG. 33 to FIG. 34, a decoding in FIG. 32, a decoder in FIG. 36, a decoding method in FIG. 38, etc.

[0242] Methods and apparatus according to embodiments include a method and apparatus on improving attribute coding and decoding of G-PCC file using geometric transformation prior region-based adaptive loop filter.

[0243] Embodiments include methods for improving the attribute representation of Geometry-Based Point Cloud Compression (G-PCC) files. The described methods are implemented based on the ISO-based media file format for G-PCC transmission.

[0244] The embodiments relate to a method for improving the compression efficiency of Geometry-based Point Cloud Compression (G-PCC) for compressing 3D point cloud data.

[0245] A point cloud consists of a collection of points, each of which can have geometry and attribute information. Geometry information includes 3D positional (XYZ) information, while attribute information includes color (RGB, YUV, etc.) and / or reflectance values.

[0246] The G-PCC encoding process can be comprised of dividing a point cloud into tiles by region, dividing each tile into slices for parallel processing, compressing geometry for each slice, and compressing attribute information based on the reconstructed geometry (decoded geometry) based on the positional information changed through compression.

[0247] The G-PCC decoding process can be configured to receive an encoded slice unit geometry bitstream and attribute bitstream, decode the geometry, and decode attribute information based on the geometry reconstructed through the decoding process.

[0248] Octree-based, predictive tree-based, or trisoup-based compression techniques can be used to compress geometry information.

[0249] For attribute information compression, compression techniques based on predicting transform, lifting transform, or RAHT transform can be used.

[0250] These embodiments relate to a technique for applying an Adaptive Loop Filter to increase the compression ratio of point cloud attributes.

[0251] Examples may refer to the document ISO / IEC 23090-9, Coded representation of immersive media - Part 9: Geometry-based point cloud compression.

[0252] A draft specification for transporting G-PCC data in the ISO Base Media File Format (ISOBMF) is in Committee Draft (CD) status.

[0253] Point clouds are widely used for several compelling reasons, including 3D representation, versatility, architecture and construction, geospatial and mapping, manufacturing and quality control, entertainment and media, medical imaging, precision and accuracy, data richness, interoperability, automation, and efficiency. Modern video coding schemes commonly incorporate loop filters in the final stage of frame or slice encoding. For example, AVC / H.264 uses a deblocking filter, while HEVC / H.265 uses both a deblocking filter and a sample adaptive offset (SAO). HEVC's deblocking filter offers two variants: weak and strong, while SAO offers two types: edge offset and band offset. VVC / H.264 adopts both an adaptive loop filter (ALF) and SAO. Furthermore, this concept is extended to cross-component filtering, exemplified by the Cross-Component Adaptive Loop Filter (CC-ALF) and cross-color filtering between layers for scalability. Determining the filter coefficients on the encoder side typically involves linear regression methods. Consequently, cross-component linear model prediction is often discussed in conjunction with loop filters, even though it does not require special filtering of neighboring pixels in this context. To enhance attribute coding efficiency, G-PCC encoders / decoders can utilize adaptive loop filters. To achieve a high level of adaptivity, flexibility, compression gain, and parallelism, adaptive loop filters are not applied just once per frame or slice, but rather as a set of different filters tailored to selected regions. The size of these regions can vary depending on conditions such as the quantization parameter (QP), sparsity, and other factors. These filters can also act as post-filters. However, if the encoder can control the ALF parameters, it can ensure the desired level of quality and pass this information to the decoder.Furthermore, the loop filter improves the quality of the reference frame, thereby increasing the efficiency of inter-prediction and reducing bit size. While post-filtering requires additional buffering, the filtered output signal can be included in the decoder picture buffer (DPB), eliminating the need for additional memory. An additional feature of the region-based adaptive loop filter is that it transmits an "on / off" flag to the decoder. This allows the encoder to determine whether to apply ALF to each small region based on rate distortion optimization (RDO) criteria or an attention (saliency) map. Finally, the loop filter improves visual quality by removing artifacts such as blocking and ringing. Therefore, this process is normative.

[0254] The definitions of terms according to the embodiments are as follows:

[0255] - Point cloud frame: a set of 3D points specified by Cartesian coordinates (x, y, z) and an optional set of corresponding fixed properties at a specific point in time.

[0256] - Bounding box: A rectangular prism containing the source point cloud frame.

[0257] - Geometry: A set of Cartesian coordinates associated with the point cloud frame.

[0258] - Attributes: Scalar or vector attributes associated with each point in the point cloud, such as color, reflectance, frame index, etc. (optional)

[0259] - APS: Attribute Parameter Set

[0260] - ASH: Attribute Slice Header

[0261] - GSH: Geometry Slice Header

[0262] - GPS: set of geometric parameters

[0263] - LSB: least significant bit

[0264] - RAHT: Domain Adaptive Hierarchical Transform

[0265] - SPS: Sequence Parameter Set

[0266] - TPS: Tile parameter set (same as tile inventory)

[0267] - Slice: A series of syntax elements representing part or all of a coded point cloud frame.

[0268] - 3D tiles: rectangular cuboids inside a bounding box

[0269] - ALF: Adaptive Loop Filter

[0270] - RDO: Rate Distortion Optimization

[0271] - HLS: High-Level Syntax

[0272] - QP: Quantization Parameter

[0273] The basic concept of the adaptive loop filter (ALF) is to minimize the mean squared error between the attributes of the original and decoded voxels through convolution using Wiener-based adaptive filter coefficients. This technique, first developed by Norbert Wiener in 1940, is applied to the final processing stage of each image to effectively remove visual artifacts from previous stages. The encoder determines the optimal filter coefficients and explicitly passes them to the decoder. To improve coding efficiency, particularly in low quantization rate (QP) scenarios, local adaptation techniques are used, applying different filters to different regions within the image. Furthermore, coding efficiency can be further improved by controlling the "on / off" of the filter in each selected region. To address this problem, the Wiener-Hoff equations must be solved in matrix form. This linear system of equations can be solved using Gaussian elimination. However, the matrix is ​​symmetric, and more efficient solutions can be obtained using specialized techniques such as Cholesky decomposition or the conjugate gradient method. Assuming coding error symmetry, it is necessary to introduce symmetric filters. These filters can reduce the number of filter coefficients by half, reducing both the number of multiplications and the overhead of transmission coefficients. Considering that multipliers occupy significantly more chip area than adders, reducing the number of multipliers significantly reduces the chip area required for ALF.

[0274] 1) The application of the region-based adaptive loop filter on the encoder side consists of the following main steps.

[0275] * Divide the point cloud into cubes of size 2p x 2p x 2p.

[0276] * Find the neighbors for the processed voxel within each cube and generate a distribution for that tab.

[0277] * Compute the matrix and dominance vector of the Wiener-Hopf equation using the original and reconstructed attribute signals.

[0278] * Find the solution to the Wiener-Hopf equation using floating-point operations.

[0279] * Scaling factor and rounding.

[0280] * Calculate convolution filter coefficients using the attribute values ​​of the referenced voxels.

[0281] * Decide whether to apply based on RDO or user saliency map.

[0282] * Write high-level syntax flags (whether ALF needs to be applied)

[0283] * Write the overhead (integer ALF coefficients) using entropy coding.

[0284] 2) The application of the region-based adaptive loop filter on the decoder side consists of the following main steps.

[0285] * Read HLS flag (whether ALF needs to be applied)

[0286] * If the ALF flag is true, read the required data from the bitstream for ALF processing for each region (2p x 2p x 2p cube).

[0287] * Divide the point cloud into 2p x 2p x 2p cubes.

[0288] * If the current area is processed with ALF, the neighbors of the processed voxels within each cube are searched.

[0289] * Calculate convolution filter coefficients using the attribute values ​​of the referenced voxels.

[0290] * For each processed voxel, neighboring voxel search and convolution calculation can be performed in parallel.

[0291] Figure 11 is an example of point cloud segmentation according to embodiments.

[0292] The method according to the embodiments comprises the following steps.

[0293] Figure 11(a) shows an example of point cloud segmentation for a 256x256x256 cube, and Figure 11(b) shows an example of point cloud segmentation for a 512x512x512 cube and cube boundaries.

[0294] Step 1) Partition

[0295] The entire region containing a given point cloud can be divided into separate cubes, each of size 2p. When the cube size is a power of 2, determining whether a given voxel is contained in a particular cube becomes straightforward. This can be done efficiently using bitwise shift operations.

[0296] If the intersection of the current cube and the given point cloud contains a sufficient number of voxels (i.e., the number of voxels exceeds a predefined threshold), the encoder generates a temporary point cloud for both the reconstructed image and the original image. This temporary point cloud consists only of voxels belonging to the intersection.

[0297] If the voxel order of the original cloud differs from that of the reconstructed cloud, the order of the temporary cloud with the original properties must be rearranged. However, in most cases, the original signal point cloud maintains the same order as the reconstructed point cloud. The size of the cube significantly affects the compression gain. A larger cube reduces the overhead of ALF data, while a smaller cube improves quality. Overhead reduction and quality improvement are often conflicting goals. For high QPs, the number of bits is more important, and overhead must be small. For low QPs, quality is more important than the number of bits. Various experiments using different QP values ​​and sequences have shown the following optimal relationship.

[0298] Table 1

[0299]

[0300] Table 1 shows the dependence of cube size on QP.

[0301] Therefore, for a size of 2p, base 2, and exponent p, this dependence can be approximated with high accuracy by the following equation: p= QP+

[0302] For Table 1, it is as follows: p=(QP-34) / 6+9

[0303] Figure 12 shows an example in which the original signal and the reconstructed signal according to the embodiments are included in one cube.

[0304] Step 2) Neighbor Search

[0305] Find the neighbors of the processed voxel within the two-dimensional cube and find the distribution for that tap.

[0306] The main goal of this step is to identify neighboring voxels and utilize their attribute values ​​in the convolution computation. This aims to minimize the difference between the reconstructed signal and the original signal using the mean squared error (MSE) as a cost function: ||Org- Filter* Rec||->min

[0307] This process presents the most complex and sensitive aspect. Simple neighborhood search, which operates as a quadratic function of the number of voxels, becomes extremely labor-intensive for a 2048-size cube. Therefore, the method according to the embodiments utilizes information from the Molten code to accelerate the search. Another issue is the neighborhood size selection criteria. The two images, Egyptian_mask_vox12 and soldier_vox10_0690, are displayed at the same scale in Figure 13.

[0308] Figure 13 shows an example of two different point clouds of the same scale, with two spheres having radii of 16 and 2 added for comparison according to the embodiments.

[0309] To balance quality improvement and overhead reduction, it is essential to use a relatively small number of taps and limit the number of neighboring voxels for each processed voxel. Therefore, the window size for determining the neighboring region where voxels are found must be selected individually for each point cloud. The window size varies significantly depending on the geometric characteristics that describe the voxel distribution within the cloud. For sparse clouds, the window size can be relatively large, while for dense clouds, it must be smaller. Therefore, the window size is directly affected by the voxel density. While voxel density can be calculated for each cube on both the encoder and decoder sides, this parameter is generally stable for most point clouds and can be determined for the entire frame or slice before being transmitted in the bitstream.

[0310] Figure 14 shows examples of Z-order curves (Lebesgue curve 1904, Morton space-filling curve 1966) according to embodiments.

[0311] Acceleration of neighboring foxel search based on Morton code information according to embodiments:

[0312] Forming filter taps requires finding all voxels in the spherical neighborhood of the current voxel. Finding these neighboring voxels (and forming a Wiener-Hopf system in general) is the bottleneck of the filtering algorithm. The computational complexity for a cube is at least O(n) (where "n" is the number of voxels occupied by the cube) and has a large multiplicative constant (consisting of neighbor candidate search, octant calculation, etc.). The solution complexity for a Wiener-Hopf system is O(t3) (where "t" is the number of filter taps, and t3 is less than or equal to n), and the multiplicative constant is very small. The complexity of applying the filter and coding / decoding the filter coefficients is O(n), which also has a small multiplicative constant. Therefore, it is important to identify the neighbors of a given voxel in constant time (which means that filter tap calculation in the encoder / decoder must be performed in O(n) time).

[0313] The proposed acceleration algorithm extracts neighbor candidates from a set of adjacent 2kx2kx2k blocks that completely contain the spherical neighborhood of the processed voxel. A block (equivalent to a cube) corresponds to a subtree of an octree. Therefore, only the 3*k low-order bits of the Morton code of a voxel within a block change, while the high-order bits remain fixed. We can assign block_code = (voxel_code >> 3*k), where voxel_code is the Morton address of an arbitrary voxel within the block, and ">>" is a bitwise right shift. A key requirement is constant-time access to the voxel list of each block, which is achieved through the following preprocessing steps.

[0314] 1) Store the occupied voxels of a slice in a sorted array using Morton code. All voxels in a block (cube) occupy consecutive slots in this array. The sorting complexity is O(N log N), where N is the number of occupied voxels in the slice. For a reasonable N, "log N" is much smaller than the multiplicative constant in the O(n) neighbor search, so we can assume O(n) complexity per cube.

[0315] 2) For each block, store the array1 location of that block in another array (array2) indexed by the block code.

[0316] So for every voxel we get a list of neighbors in the same block in constant time: array1[array2[voxel code >> 3*k]].

[0317] A block is defined by the voxel codes of a suitably chosen set of nearby voxels. Examples include (a spherical neighborhood is defined as ||neigh - vox|| < R):

[0318] a) Eight corner voxels of a cubic neighborhood (see Figure 6). The block size is chosen such that k >= 2*R - 1 (to ensure that the blocks are contiguous).

[0319] b) 27 voxels: corners + middle corners + center (Rubik's Cube). K >= R - 1.

[0320] Option (b) requires exploring approximately twice as many candidate voxels as option (a). However, for a reasonable R neighborhood, there are only a few dozen voxels. Therefore, the overhead of selecting 27 - 8 = 19 additional blocks (including duplicate removal) experimentally overweights this twofold reduction in voxel count, making option (a) computationally cheaper.

[0321] Figure 15 shows an example of a 2D projection of a block defined by 8 corner voxels and 27 corner + middle corner + center voxels according to embodiments.

[0322] Memory requirements for neighboring voxel search according to embodiments:

[0323] Configuring array2 for the entire slice results in a very large memory footprint. The size of array2 is approximately equal to the number of blocks in the slice geometry (2kx2kx2k), and is no smaller than the number of blocks_codes between the block_code of the first and last occupied blocks in the slice. This number can be very large. However, this approach can slightly improve compression ratio by using taps outside the cube for filtering.

[0324] Figure 16 shows an example of retrieving truncated_block_code from the Morton address of a voxel according to embodiments.

[0325] The method according to the embodiments further includes configuring array2 within a cube to reduce memory requirements and indexing elements using truncated_block_code. In this approach, the size of array2 is limited to 23*(pk), which is not large compared to an appropriate cube size. Only tabs within the cube can be used for filtering. The cube can be processed in parallel:

[0326] Truncated_block_code = (voxel_code & ((1 << 3*p) - 1)) >> 3*k,

[0327] Here, & is bitwise AND, << is bitwise left shift, and >> is bitwise right shift.

[0328] Figure 17 shows information about a Molton code for neighbor search according to embodiments.

[0329] Neighboring voxel distribution for the corresponding filter tap according to the embodiments:

[0330] The neighboring voxel distribution for the filter tap according to the embodiments is processed according to the following principles: 1. Distance, 2. Direction

[0331] Figure 18 shows the width distribution of filter taps according to distance (for 2D and 3D cases) according to embodiments.

[0332] 1. Distance:

[0333] According to this rule, the attribute values ​​of the reconstructed voxel are distributed to different filter taps according to the distance between the neighboring voxels and the center (processed voxel).

[0334] Figure 18 shows three regions for the neighboring voxel distribution. In this example, the four-tap filter can be thought of as follows.

[0335] filter = A1value1 + A2value2 + A3value3 + A4

[0336] value1 is the attribute value of the processed voxel, and value2 is calculated using the attributes of all voxels inside the small sphere.

[0337] value3 is calculated using all voxel properties inside the large sphere and outside the small sphere, and A4 is an offset. There is no guarantee that voxels exist in the calculated values1 and value2, and in this case, the default value is 0.

[0338] The distance between voxels can be any function that defines a metric in 3D space.

[0339] Example: abs(dx)+ abs(dy)+ abs(dz) or sqrt(dx2 + dy2+dz2)

[0340] Fig. 19 shows the filter shape of HM.7 (reference s / w of HEVC) according to embodiments.

[0341] 2. Direction:

[0342] direction.

[0343] To better adapt to the local shape of a given point cloud, different taps can be assigned to different directions. However, ALF development for modern video codecs makes a simplifying assumption about the behavior of the autocorrelation function: it is symmetric about the center.

[0344] Figure 20 shows the width distribution of filter taps according to the direction according to the embodiments (3D case).

[0345] This leads to the obvious conclusion that the filter coefficients for opposite points are identical, reducing the overhead by half. According to this rule, the attribute values ​​of a reconstructed voxel are distributed across different filter taps according to the direction from the center (the processed voxel) to neighboring voxels, but voxels with opposite directions are combined into the same region. The eight octant (orientation) distributions for neighboring voxels are shown in Figure 20. Symmetric octants are combined into a single region.

[0346] In this example, we can consider six tap filters.

[0347] Filter = A0value0 + A1value1 + A2value2 + A3 value3 + A4 value4 + A5

[0348] Here, value0 is the attribute value of the processed voxel. value1, value2, value3, value4, and value5 are calculated using all voxel attributes within each octant, and A5 is an offset.

[0349] Figure 21 shows an example of a 4-tap symmetric ALF design using 8 octants according to embodiments.

[0350] If the attribute values ​​of neighboring voxels are the same as or very close to those of the processed voxel, the condition number of the matrix in the Wiener-Hopf equation may be high. This means that the problem becomes an ill-posed problem. Instead of using the attributes when calculating the values, it is better to calculate the attribute differences between neighboring voxels and the processed voxel. This technique improves the situation when the solution of the Wiener-Hopf equation is an integration problem and significantly reduces the dynamic range of intermediate calculations. Let Y0 be the attribute value of the processed voxel and Yk (k=1, N1) be the attribute values ​​of neighboring voxels belonging to two symmetric octants 1 and 5 (Fig. 21). Therefore, the set Yk - Y0 is used to calculate value1. To determine value1, the median of {Yk - Y0} or the most frequently occurring value in the data set can be used.

[0351] value 1=med{Y_k-Y_0} or value1=mode{Y_k-Y_0}

[0352] When Yk-Y0 are integer values, the median and mode are also integers.

[0353] Another possibility is the mean. This is very easy to find, but when Yk-Y0 are integers, the mean is very likely not to be an integer.

[0354] Apply filters:

[0355] Filter = A0value0+A1value1+ A2value2+A3value3+A4value4+ A5

[0356] The results obtained by using the difference between value1, value2, value1, and value1 show that in most cases, A0 is very close to "1" and A5 is very close to "0". This is helpful for designing most simple 4-tap filters.

[0357] Filter = value0+A1value1+ A2value2+A3value3+A4value4

[0358] The process of calculating the matrix and the column vector of the Wiener-Hopf equation using the original and reconstructed attribute signals according to the embodiments is as follows:

[0359] Without losing generality, the filter has the following form: The last value, even if an offset is required. N is equal to 1.

[0360]

[0361] The unknown coefficient {coeffn} can be derived from the minimum of the squared errors.

[0362]

[0363] In this formulation, the problem is linear regression. The necessary conditions for obtaining a minimum are derived by differentiating with respect to the unknown coefficients, resulting in a linear system called the Wiener-Hoff equations.

[0364]

[0365] Here A=V T V={ }, and B= V T Org ={ }, V={value_in}, Org={ }, C={ } is the same.

[0366]

[0367] In most cases, the matrix is ​​symmetric and positive definite. However, if the determinant of matrix (A) is zero, the problem becomes uncertain, and ridge regression becomes a more desirable approach.

[0368]

[0369] Here, ε>0 is a small positive integer and I is the same matrix. This not only solves the integration problem, but also reduces the vector length and the number of bits required for the coefficients.

[0370] Solving a system of linear equations (reference). This is an encoder-only option. While Gaussian elimination is always possible, the symmetric and positive definite nature of the matrix makes Cholesky decomposition a much faster and more stable alternative. This technique is also known as the square root method.

[0371] The Cholesky decomposition for a 3x3 matrix is ​​given by the following formula:

[0372]

[0373] Scaling factors and rounding operations according to embodiments:

[0374] Numerous experiments have shown that the coefficient value1 is close to "1" and the offset is close to "0." Therefore, to reduce overhead, we use these values ​​as default values. To reduce correlation, we use deltak = (valuek -value0).

[0375] The final design is as follows:

[0376]

[0377] Matrices A and B are as follows:

[0378]

[0379] The filter coefficients are derived by solving the Wiener-Hopf equation. The resulting filter coefficients are in floating-point format, but representing variables as floating-point values ​​is particularly challenging for hardware design. Since the fractional part of the filter coefficients is limited to 2k bits, the derived filter coefficients must be scaled to this precision and rounded. The scaling factor QSTEP is chosen to be 2k = 2, 4, 8, 16, 32, etc., to allow shift operations instead of division.

[0380] Figure 22 shows a flowchart of resizing and rounding according to embodiments.

[0381] Filter coefficient limits according to embodiments:

[0382] Experimental results show that most coefficients before scaling fall in the interval [-0.5, 0.5]. To prevent integer overflow and reduce overhead, filter coefficient constraints are introduced. On the encoder side, constraints on the absolute values ​​of coefficients are needed. This can be achieved by using naive clipping or another solution with a higher ε, (A+εI)C=B. If it is known in advance that all unscaled coefficients fall in the interval [-0.5, 0.5], then integer coefficients fall in the interval [-QSTEP / 2, QSTEP / 2]. With this priori information, overhead can be significantly reduced by using truncated unary codes or Rice codes.

[0383] Calculation of convolution filter coefficients using attribute values ​​of reference-processed voxels according to embodiments:

[0384] The application of the adaptive loop filter to each voxel is independent and can be performed in parallel. Assuming that QSTEP is chosen as 2k.

[0385] And Icoeff is the scaled and rounded coefficient corresponding to the coefficient.

[0386]

[0387] Since Qstep is a power of 2, the final operation " / " can be performed using a shift operation (>>k). The clipping operation ensures that attribute values ​​do not fall outside the allowed range. The Qstep coefficients can be -1, 0, or 1, which means that filters can be designed without multiplication.

[0388] Figure 23 shows a flowchart of calculating ALF tap values ​​according to embodiments.

[0389] ALF process using average values ​​in integer operations according to embodiments:

[0390] The input to this process is an array of differences between the processed voxels and their neighbors before the adaptive loop filter DELTA_L. The output of this process is an array of reconstructed image samples modified after the adaptive loop filter alfFilterL.

[0391]

[0392] To avoid wasteful division, delta k is DELTA k =2 n delta k can be replaced with

[0393]

[0394]

[0395] For example, N from 0 to 10 k for In the case of , in the above formula is a lookup table mult(N k ) can be replaced with.

[0396] mult

[0011] = { 0, 1024, 512, 341, 256, 204, 170 , 146, 128, 113, 102}

[0397] Therefore, in the ALF process, it is necessary to store the sum of the attribute values ​​and the number of adjacent voxels.

[0398] Figure 24 shows the dynamic range for calculating a four-tap filter according to embodiments.

[0399] Dynamic range of 4-tap filter calculations using mean values ​​in integer operations according to embodiments:

[0400] QSTEP is 2 k is selected as DELTA_k= Assuming delta_k and the attributes are in [0, 2m-1], all calculations correspond to 2+k+n+m bits. For example, QSTEP=64, k=6, 2^n=1024, n=10, the attributes correspond to [0, 255], and m=8. All calculations correspond to a 26-bit buffer.

[0401] Application decisions based on RDO or saliency map according to embodiments:

[0402] The coding efficiency of ALF applications can be improved by determining Rate Distortion Optimization.

[0403] In terms of encoding, SSD norms can be pre-calculated before filtering.

[0404]

[0405] After filtration,

[0406]

[0407] The ratio distortion optimization decision can be made based on the following simple conditions:

[0408]

[0409] Here, "bits" is the number of bits required for the current cube's overhead. This number can be estimated with high accuracy using a real-world model or approximated using only integer AlF coefficients. If the condition is true, it means that additional overhead is required to improve quality and a filter should be applied. Parameter The optimal value of follows the following law:

[0410]

[0411] However, this constant varies depending on the color component, slice type, sparsity, etc.

[0412] Figure 25 illustrates an RDO-based application of ALF for various QPs according to embodiments.

[0413] As shown in Fig. 25, it can be seen that ALF is applied to target areas according to ALF activation.

[0414] The subjective improvement of ALF application can be confirmed through the determination of saliency maps. It is well known that viewers do not pay attention to the entire image, but rather focus on small, specific areas.

[0415] Figure 26 shows examples of the most interesting parts within the images according to embodiments.

[0416] A saliency map can be found at the encoder's end using an algorithm or provided by the user. ALF reduces coding errors in salient regions, contributing to improved subjective quality.

[0417] The encoding method according to the embodiments can encode point cloud data and generate related parameters (which may be referred to as signaling information, metadata, etc.). The encoding method according to the embodiments can generate a bitstream including the point cloud data and parameters. The decoding method according to the embodiments can decode the point cloud data within the bitstream based on the parameters within the bitstream.

[0418] Figure 27 illustrates the attribute parameter set data unit syntax in a bitstream according to embodiments.

[0419] High-level syntax flags:

[0420] During the attribute information encoding / decoding process, adaptive loop filter (ALF) related syntax can be added to the attribute parameter set and attribute data unit for signaling.

[0421] Parameters specified by an APS apply to all DUs where that APS is enabled.

[0422] aps_attr_parameter_set_id identifies the APS so that other DUs can reference it.

[0423] aps_seq_parameter_set_id identifies the active SPS with sps_seq_parameter_set_id.

[0424] attr_coding_type specifies the attribute coding method. Valid values ​​are listed in the table below.

[0425] Information according to the value of attr_coding_type is as follows.

[0426]

[0427] attr_primary_qp_minus4 + 4 represents the QP of the primary attribute component before adding offsets per slice, per region, and per transform level.

[0428] attr_secondary_qp_offset represents the offset to be applied to the primary attribute QP to derive the QP of all secondary attribute components.

[0429] attr_qp_offsets_present indicates whether the per-slice attribute QP offsets attr_qp_offset[ ] are present in the ADU header (if 1) or not (if 0).

[0430] aps_extension_present indicates whether the APS syntax structure contains the aps_extension_data syntax element. In bitstreams conforming to this version of this document, aps_extension_present is 0. The value 1 for aps_extension_present is reserved for future use in ISO / IEC.

[0431] aps_extension_data can have any value. The presence and value of this parameter do not affect the decoder's conformance to the profiles specified in this version of this document. The decoder ignores all aps_extension_data syntax elements.

[0432] attr_adaptive_loop_filter_enabled indicates whether attribute coding is applied using an adaptive loop filter (if 1) or not (if 0). If attr_adaptive_loop_filter_enabled is not present, it is inferred to be 0.

[0433] attr_alf_component_type indicates the attribute component to which the adaptive loop filter is applied. Valid values ​​are listed in the table below.

[0434]

[0435] Figure 28 illustrates the attribute data unit syntax in a bitstream according to embodiments.

[0436] An ADU encodes the attribute value for a single attribute within a slice. An ADU consists of an ADU header and either attribute coefficients (attribute_coeffs) (if transform coding is enabled) or directly encoded attribute values ​​(attribute_raw).

[0437] When transform coding and adaptive loop filter are enabled, the attribute ALF coefficients (attribute_alf_coeffs) consist of an ADU header and attribute coefficients.

[0438] Figure 29 illustrates the attribute data unit header syntax in a bitstream according to embodiments.

[0439] adu_attr_parameter_set_id indicates the active APS with aps_attr_parameter_set_id.

[0440] adu_temporal_id specifies the temporal ID of the frame associated with the attribute data unit.

[0441] adu_sps_attr_idx identifies an attribute coded as an index into the list of active SPS attributes. At the beginning of every ADU, the variable AttrIdx is set to adu_sps_attr_idx.

[0442] AttrIdx = adu_sps_attr_idx

[0443] If attr_coding_type is not 3, an ADU-coded attribute has at most three components.

[0444] adu_slice_id represents the slice_id value of the previous GDU.

[0445] attr_alf_window_size indicates the ALF window size.

[0446] attr_alf_qstep_log2 represents the ALF coefficient quantization step.

[0447] attr_alf_geom_shift represents the geometric coordinate shift value for ALF.

[0448] Predicting filter coefficients of already processed blocks according to embodiments:

[0449] Coefficients in adjacent regions exhibit a high correlation. In particular, in 70% of cases, coefficients improve the quality of not only the current region but also the next region. Therefore, filter coefficient prediction can reduce overhead.

[0450] Methods and devices according to embodiments encode and decode attribute data based on region-based ALF for reflectance and scant content.

[0451] Truncated Rice codes according to embodiments:

[0452] A truncated Rice code is used to encode the absolute values ​​of the filter coefficients. Geometric distribution The optimal rice parameter n can be obtained from the following equation.

[0453]

[0454] Scaling factor QSTEP= If is finally considered

[0455]

[0456]

[0457] This formula assumes a geometric distribution and provides a rough estimate of the Rice parameter, bypassing coding. However, it explains why the Rice parameter should increase as the scaling factor increases.

[0458] Encoding overhead (such as integer ALF coefficients and ALF flags) is handled using entropy coding. The encoder / decoder already knows how many cubes are used for each slice / frame. First, the encoder determines whether it needs to write an ALF on / off flag to each cube. If the encoder determines that this flag is too expensive, the coefficients are written as 0 when ALF is not applied to the current cube.

[0459] Overhead for luma according to embodiments (integer ALF coefficients):

[0460] The pseudocode for encoding attribute filter coefficients is as follows:

[0461] void encodeCoeffALF (vector <int>& flag, vector <int>& coeff,

[0462] PCCResidualsEncoder& encoder,bool filter_apply) {

[0463] if(filter_apply == false) return;

[0464] int n = flag.size();int index_coef = 0;

[0465] bool need_to_write_flag=check_need_to_write_flag(nuber_of_zeros, n);

[0466] for (int i = 0; i < n; i++){

[0467] if (need_to_write_flag) { encodeALFflag(flag[i]);

[0468] if (flag[i]) encodecoeffLast());

[0469] }else {

[0470] if (flag[i]){ encodecoeff();

[0471] } else {encodezeros();

[0472] }

[0473] Figure 30 illustrates a flowchart of ALF coefficient coding according to embodiments. The method according to embodiments: Start the encodeCoeffALF function. Check whether filter_apply is false. If false, the filter is not applied, so the process ends. Check the number of zeros. If the filter needs to be applied, the check_need_flag(number_of_zeros) function is called. This determines the number of zero coefficients and determines whether special processing is required. It repeats while the index i is less than n, and is divided into two paths. 1) Zero coefficient processing: If the flag is false, the encodezeros function is called to efficiently encode consecutive zero coefficients. Then, if the flag becomes true, the encodecoeff function is called to encode the actual coefficient value. The process ends when i >= n. 2) ALF flag processing: If need_flag is true and i < n, the ALF flag is processed. If the flag is false, the process ends immediately. If the flag is true, the encodecoeffLast function is called to specially encode the last coefficient.

[0474]

[0475] Fig. 31 shows a flowchart of ALF coefficient coding for four-tap filters encodeCoeff, encodeCoeffLast, and encodeCoeffzeros according to embodiments.

[0476] Coeff encoding process: Sequentially encode four coefficients from coeff[index + 0] to coeff[index + 3]. Each coefficient is encoded using Rice coding by calling the encodeALFcoeffRice() function. After processing all coefficients, the index is increased by 4 (index = index + 4).

[0477] CoeffLast encoding process: Encode the first three coefficients (coeff[index + 0], coeff[index + 1], coeff[index + 2]) normally. Check the needLast flag in the fourth coefficient (coeff[index + 3]). If needLast is true: Encode the last coefficient in a special way using the RiceLast() function. If needLast is false: Use the encodeALFcoeffRice() function. After processing, increase the index by 4.

[0478] encodeZeros: This function efficiently handles consecutive zero coefficients. Call encodeALFcoeffRice(0) four times in a row. This encodes that all coefficients are zero. Instead of encoding zero values ​​individually, it processes them in a group, improving compression efficiency.

[0479] If the flag is set and the first three coefficients are 0, the last coefficient is definitely not 0, so encodeCoeffLast can be applied to reduce the absolute value of the coefficient by 1. If the flag is not set, encodezeros is used.

[0480] For a 4-tap filter, nine context models are required for overhead encoding: one context model for the flag (ctxALF[0]), four context models for the absolute value (ctxALF[1], ctxALF[2], ctxALF[3], ctxALF[4]), and four context models for the sign of each coefficient (ctxALF[5], ctxALF[6], ctxALF[7], ctxALF[8]).

[0481] The truncated Rice coding for the ALF coefficient pseudocode is:

[0482] void encodeALFcoeffRice ( int coeff, int index, int max, int Riceparameter, arithmeticEncoder) {

[0483] Level =abs(coeff);

[0484] Level1= Level>>Riceparameter; int max1 = max >> Riceparameter;

[0485] max1 = max >> Riceparameter; reminder = level - (level1 << Riceparameter)

[0486] reminder = level - (level1 << Riceparameter);

[0487] for ( i = 0; i < level1; i++){

[0488] arithmeticEncoder.encode(0, ctx[index+1]);

[0489] if (level1 < max1 - 1)

[0490] arithmeticEncoder.encode(1, ctx[index+1]);

[0491] for ( i = 0; i < Riceparameter; i++){

[0492] simbol = reminder & 1;

[0493] arithmeticEncoder.encode_by_pass(simbol);

[0494] if (level > 0){

[0495] reminder =reminder >>1;}

[0496] if (level > 0){

[0497] arithmeticEncoder.encode(sign, ctx[5+ index]);}

[0498] Figure 32 shows a flowchart of an ALF coefficient decoding process according to embodiments.

[0499] The method of Fig. 32 is the ALF coefficient decoding of the decodeCoeffALF function, and operates in the reverse order of the encoding process of Figs. 30 and 31:

[0500] The decodeCoeffALF function begins. First, it calls the decodeALFflag() function to decode whether a filter is applied. If filter_apply is false, the function terminates immediately, as no filter is used. Flag information is decoded. If a filter is applied, the decodeALFflag() function is called again to decode additional flag information. The value of need_flag is checked. 1) Coefficient decoding: If need_flag is false, follow this path. Repeat while i < sizeflag is satisfied. Decode the coefficients by calling the decodeALF() and Checkcflag() functions. Increment the index i and continue processing. 2) Flag-based decoding: If need_flag is true, follow this path. While i < sizeflag: Call the flag[i] = decodeALFflag() function to decode the flag at each position. Check the decoded flag value. If flag is true: Call the decodeALFLast() function to specifically decode the last coefficient. If flag is false: the coefficient is skipped.

[0501] The syntax and semantics for encoding and decoding according to the embodiments are as follows. The attribute data unit according to the embodiments further includes syntax element(s) for coded ALF coefficients.

[0502] Figures 33a and 33b illustrate attribute data unit ALF coefficient syntax according to embodiments.

[0503] Figure 34 illustrates the attribute ALF coefficient tuple syntax according to embodiments.

[0504] Attribute Data Unit Semantics:

[0505] An ADU encodes the attribute value of a single attribute within a slice. An ADU consists of an ADU header, attribute coefficients (attribute_coeffs) or directly encoded attribute values ​​(attribute_raw) if transform coding is enabled, and in-loop filter coefficients if filtering is enabled in the ADU header.

[0506] As a syntax related to slice attributes, the bitstream may further include syntax elements related to encoded ALF coefficients.

[0507] Attribute Data Unit ALF coefficient related parameters:

[0508] The AttrALFCoeff array contains AttrALFCoeff[cubeidx ][*?*t*?*][*?*tapsidx*?*] elements, which contain the scaled filter coefficients. The elements of the array are initialized to 0. AttrALFCoeff contains NumALFTaps filter coefficients per attribute channel per cube. The first NumALFTaps-1 coefficients are coded in the attribute_alf_coeff_tuple() function. The last filter coefficient is for the center (filtered) voxel and is inferred to be 1. If all filter coefficients coded in the attribute_alf_coeff_tuple() function are 0, ALF is not applied to the cth channel of the corresponding cubeidxth cube.

[0509] adu_alf_flag indicates whether ALF filter coefficients are present in the ADU (i.e., whether ALF is applied to reconstructed voxels belonging to that ADU).

[0510] If alf_flag_cube_present is 1, the filter application for each cube is explicitly indicated via the alf_flag_cube, alf_flag_cube_cb, and alf_flag_cube_cr flags. If alf_flag_cube_present is 0, the ALF coefficients for each cube are transmitted.

[0511] Attribute ALF coefficient tuple related parameters:

[0512] According to zeroCoeffs, alf_coeff_abs[ tapsidx], last_alf_coeff_abs_minus1, and alf_coeff_sign[ tapsidx] together represent the tapsidx-th filter coefficient ALFAttrCoeff[ cubeidx*?*][*?* ]. coeff_sign[ ] indicates whether the sign of the coefficient is positive when it is 0 or negative when it is 1. When the filter coefficients of the tuple 0, 쪋, and NumALFTaps-2 are 0, zeroCoeffs is equal to 1.

[0513] The coefficients of the cubeidxth cube and the th attribute component are specified through the derivation of ALFAttrCoeff.

[0514] zeroCoeffs = 1

[0515] for (tapsidx = 0; tapsidx < NumALFTaps - 1; tapsidx++){

[0516] absVal = alf_coeff_abs[tapsidx]

[0517] if (absVal)

[0518] zeroCoeffs = 0

[0519] if (tapsidx == NumALFTapsMinus1- 1 && zeroCoeffs)

[0520] absVal = last_alf_coeff_abs_minus1 + 1

[0521] ALFAttrCoeff[cubeidx][c][tapsidx] = (1 - 2 Х alf_coeff_sign[tapsidx]) Х absVal

[0522] ALFAttrCoeff[cubeidx][c][NumALFTaps - 1] = 1

[0523] With respect to slice attributes, the following embodiments can be defined as attribute filtering using an adaptive loop filter.

[0524] The method according to the embodiments further includes an attribute filtering step using an adaptive loop filter. The attribute filtering step may include a filtering preparation step, a filter coefficient order setting step, and an attribute filtering step.

[0525] The adaptive loop filter (ALF) is a FIR in-loop filtering technique that aims to minimize the MSE distance between the reconstructed attribute plane and the original attribute plane. This filter is applied when attr_alf_enabled is 1. The encoder solves the Wiener-Hopf equation using floating-point arithmetic to obtain the filter coefficients. The decoder reads the filter coefficients / kernel from the ADU, computes the filter taps on the reconstructed attribute plane, and performs a convolution of the computed taps with the filter kernel, all using integer arithmetic. The filtering process takes the reconstructed attribute coefficient PointAttr as input and outputs the filtered attribute coefficient FilteredPointAttr.

[0526] Prepare to filter:

[0527] The filtering process is specified for a single voxel. The preparation phase consists of:

[0528] - Read ALF coefficient quantization step ALFQstep = Exp2(attr_alf_qstep_log2), filter window size ALFWin = attr_alf_window_size, geometric coordinate shift ALFGeomShift = attr_alf_geom_shift

[0529] - Expands the geometric coordinates by moving the voxel coordinates. This movement may create additional duplicate points, but does not affect the filter application process.

[0530] for (ptIdx = 0; ptIdx < PointCnt; ptIdx++)

[0531] for (k = 0; k < 3; k++) {

[0532] ALFAttrPos[ptIdx][k] = AttrPos[ptIdx][k] >> ALFGeomShift}

[0533] - Determine cube size (log, ALFLog2CubeSize):

[0534] ALFLog2CubeSize := 9 + (qpP - 34) / 6

[0535] - Iterate over all cubes in the slice (cube size is 1 << ALFLog2CubeSize) and count the number of cubes that contain more voxels than ALFCubeVoxelThreshold to calculate the number of populated cubes (NumPopulatedCubes).

[0536] Filter coefficient order:

[0537] The filter is applied if NumPopulatedCubes > 0 and alf_flag == 1. The filter coefficients are then decoded (the ALFAttrCoeff array is created). ALFAttrCoef contains filter coefficients only for filled cubes. cubeidx enumerates the filled cubes in ascending order of ALFCubeCode.

[0538] ALFCubeCode = Morton[ALFAttrPos[cubePtIdx]] >> 3 * ALFLog2CubeSize

[0539] Here, cubePtIdx is the index of all voxels belonging to the cube.

[0540] Attribute filtering:

[0541] Filtering is applied independently to each cube. Filter taps in adjacent cubes are removed by the decoder.

[0542] Calculating the Filter tab of Attribute Filtering:

[0543] The filter taps are computed using the attribute values ​​of voxels in the spherical neighborhood of the voxel. For the current voxel with ptIdx = curPtIdx, the candidate neighboring voxels with ptIdx = candPtIdx satisfy the following conditions:

[0544] (ExSquared[ALFAttrPos[curPtIdx][0] - ALFAttrPos[candPtIdx][0]] + ExSquared[ALFAttrPos[curPtIdx][1] - ALFAttrPos[candPtIdx][1]] + ExSquared[ALFAttrPos[curPtIdx][2] - ALFAttrPos[candPtIdx][2]]) < ExSquared[ALFWin]

[0545] To compute the filter taps, the decoder identifies all these neighboring voxels.

[0546] The current filter design implies NumALFTaps = 5. Taps with tapsidx = 0, 1, 2, and 3 are defined by eight octants of a spherical neighborhood. Voxels in opposite octants contribute to the same tap. For each neighboring voxel, a tap index is assigned using the following logic:

[0547] dk0Gt0 = ALFAttrPos[curPtIdx][0] - ALFAttrPos[candPtIdx][0]

[0548] dk1Gt0 = ALFAttrPos[curPtIdx][1] - ALFAttrPos[candPtIdx][1]

[0549] dk2Gt0 = ALFAttrPos[curPtIdx][2] - ALFAttrPos[candPtIdx][2]

[0550] if ((dk0Gt0 == true && dk1Gt0 == true && dk2Gt0 == true) || (dk0Gt0 == false && dk1Gt0 == false && dk2Gt0 == false))

[0551] tapsidx = 0;

[0552] else if ((dk0Gt0 == false && dk1Gt0 == true && dk2Gt0 == true) || (dk0Gt0 == true && dk1Gt0 == false && dk2Gt0 == false))

[0553] tapsidx = 1;

[0554] else if ((dk0Gt0 == false && dk1Gt0 == false && dk2Gt0 == true) || (dk0Gt0 == true && dk1Gt0 == true && dk2Gt0 == false))

[0555] tapsidx = 3;

[0556] else

[0557] tapsidx = 4;

[0558] The filter tap for the c-th attribute at the curPtIdx-th voxel is described by the derivation of ALFTaps. When tapsidx = 0, 쪋, 3, ALFTaps[curPtIdx][c][tapsidx] is equal to the difference between the attribute value of the current voxel and the arithmetic mean of the attribute values ​​of the neighboring voxels whose tap indices are tapsidx. Then ALFTaps[curPtIdx][c][NumALFTaps - 1] = PointAttr[curPtIdx][c]. The representation of ALFTaps requires floating-point operations.

[0559] The decoder only performs integer arithmetic. Therefore, the filter tap for the c-th attribute of the curPtIdx-th voxel is specified by deriving ALFTapsSumAttributes and ALFTapsNumPt. Here, ALFTapsSumAttributes[c][tapsidx] is the sum of the attribute values ​​of the neighboring voxels whose tap indices are tapsidx, and ALFTapsNumPt[tapsidx] (same for all c) is the number of such neighboring voxels. The arithmetic mean calculation is integrated into the filter coefficient scaling process, where division is replaced by multiplication and shift operations.

[0560] Applying filters for attribute filtering:

[0561] The inputs to this process are the reconstructed attribute coefficients PointAttr[curPtIdx][c], the filter taps ALFTapsSumAttributes[c], ALFTapsNumPt, the decoded filter coefficients ALFAttrCoeff[cubeidx*?*][*?*], and the scaling factor ALFQstep. The output is the filtered reconstructed coefficients FilteredPointAttr[curPtIdx][c].

[0562] The decoded filter coefficients ALFAttrCoeff are scaled versions of the final filter coefficients with the scaling factor ALFQstep. The effect of the division AttrALFCoeff[cubeidx ][*?* ][*?*tapsidx*?*] / ALFQstep in the coefficient computation and the effect of the division ALFTapsSumAttributes[c][tapsidx] / ALFTapsNumPt[tapsidx] in the tap computation are combined into a single integer calculation. ALFQstep is equal to a power of 2.

[0563] Fixed-point arithmetic using bit precision ALFBitPrecision is used.

[0564] Integer array ALFMult:

[0565] const int ALFMult

[0101] = {

[0566] 0, 1024, 512, 341, 256, 204, 170, 146, 128, 113, 102, 93, 85, 78, 73,

[0567] 68, 64, 60, 56, 53, 51, 48, 46, 44, 42, 40, 39, 37, 36, 35,

[0568] 34, 33, 32, 31, 30, 29, 28, 27, 26, 26, 25, 24, 24, 23, 23,

[0569] 22, 22, 21, 21, 20, 20, 20, 19, 19, 18, 18, 18, 17, 17, 17,

[0570] 17, 16, 16, 16, 16, 15, 15, 15, 15, 14, 14, 14, 14, 14, 14, 13,

[0571] 13, 13, 13, 12, 12, 12, 12, 12, 12, 12, 12, 12, 11, 11, 11,

[0572] 11, 11, 11, 10, 10, 10, 10, 10, 10,

[0573] };

[0574] Contains precomputed 2ALFBitPrecision / NumPt values ​​for all possible NumPt - voxel counts in the tab.

[0575] ALFMult[ALFTapsNumPt[tapsidx]] = (1 << ALFBitPrecision) / ALFTapsNumPt[tapsidx]

[0576] The filter application process forms a FilteredPointAttr array, which is defined as follows:

[0577] rightshift = ALFBitPrecision + log2(ALFQstep)

[0578] for (tapsidx = 0; tapsidx < NumALFTaps - 1; tapsidx++){

[0579] delta = ALFTapsSumAttributes[tapsidx] - PointAttr[curPtIdx][c]*ALFTapsNumPt[tapsidx]

[0580] term_scaled = ALFAttrCoeff[cubeidx][c][tapsidx]*delta*ALFMult[ALFTapsNumPt[tapsidx]]

[0581] term = (term_scaled*Sign(term_scaled) + (1 << (rightshift - 1)) >> rightshift

[0582] term[tapsidx] = term*Sign(term_scaled)

[0583] FilteredPointAttr[curPtIdx][c] = PointAttr[curPtIdx][c] + term[0] + term[1] + term[2] + term[3]

[0584] Fig. 35 shows an encoding device according to embodiments.

[0585] The device of Fig. 35 is an encoding device according to embodiments and follows the description of Fig. 1 transmitting device (10000), encoder (10002), Fig. 2 acquisition (20000) to transmission (20002), Fig. 4, Fig. 12 encoder, Fig. 10 each device, Fig. 11 to Fig. 26 ALF (Adaptive Loop Filter) operation, Fig. 27 to Fig. 29, Fig. 33 to Fig. 34 HLS (High Level Syntax) generation, Fig. 30 to Fig. 31 encoding, Fig. 37 encoding method, etc.

[0586] Each component constituting the block diagram of the PCC data encoder of Fig. 35 may correspond to hardware, software, a processor, and / or a combination thereof. The input of the decoder includes an encoded geometry information bitstream and an attribute information bitstream, which are decoded to output restored PCC data.

[0587] Data Input: Receives geometry, attributes, and parameters. Coordinate Transformation: Converts the coordinates of the input point cloud into a form suitable for processing. Geometry Information Transformation Quantization Processing: Quantizes 3D coordinate information to improve compression efficiency.

[0588] Geometry encoding related: Voxelization processing: Convert the point cloud into voxels (3D pixels). Frame classification: Process by distinguishing between P-frames and I-frames. Geometry information intra-prediction: Predict geometry information within the current frame. Geometry information inter-prediction: Predict by comparing with the previous frame. Final frame generation: Generate the final frame based on the predicted information.

[0589] Attribute encoding related: Color Transform Processing: Converts RGB color information into a form suitable for compression. Lossy-Geom Decision: Determines whether lossy compression is used. For example, if Yes, recoloring is performed. If No, the next step is performed. Attribute Inter-prediction: Performs inter-prediction of attribute information. Attribute Intra Prediction: Performs intra-prediction of attribute information. Apply Adaptive Loop Filter (ALF): Performs filtering to reduce compression artifacts.

[0590] Reconstructed Geometry: Reconstructs geometry information (geometry data). Entropy Encoding: Encodes the final compressed data using entropy coding. The encoding outputs are a geometry bitstream and an attribute bitstream.

[0591] Fig. 36 shows a decryption device according to embodiments.

[0592] The device of Fig. 36 is a decoding device according to embodiments, and follows the description of the receiving device (10004) of Fig. 1, the decoder (10006), the transmission (20002) to the rendering (20004) of Fig. 2, the decoder of Fig. 7, Fig. 13, each device of Fig. 10, the ALF (Adaptive Loop Filter) operation of Figs. 11 to 26, the HLS (High Level Syntax) acquisition of Figs. 27 to 29, and Figs. 33 to 34, the decoding of Fig. 32, and the decoding method of Fig. 38.

[0593] Each component constituting the block diagram of the PCC data decoder of Fig. 36 may correspond to hardware, software, a processor, and / or a combination thereof. The input of the decoder includes an encoded geometry information bitstream and an attribute information bitstream, which are decoded to output restored PCC data.

[0594] The Fig. 36 decoder can follow the reverse process of the operation of the Fig. 35 decoder.

[0595] The decoder receives compressed geometry information (Geometry Bitstreams) and compressed attribute information (Attribute Bitstreams).

[0596] Regarding geometry decoding, geometry information entropy encoding: Decodes the compressed geometry bitstream. Determines whether intra coding is performed. For example, if Yes, performs geometry information intra prediction restoration, and if No, performs geometry information inter prediction restoration. Reference frame generation: Creates a reference frame using the restored geometry information. Geometry information transformation dequantization processing: Restores quantized data to the original. Coordinate inverse transformation: Restores the transformed coordinates to the original coordinate system.

[0597] Regarding attribute decoding, attribute information entropy encoding: Decode the compressed attribute bitstream. Determine whether intercoding is performed. For example, if yes, perform Attribute Inter Prediction prediction restoration, and if no, proceed with Attribute Intra Prediction prediction restoration.

[0598] Determines whether to use an adaptive loop filter. For example, if Yes, compression artifacts are removed in the ALF application section; if No, the process proceeds directly to the next step.

[0599] Through decoding, restored attribute information is output. The original color is restored by performing the reverse process of color conversion.

[0600] As the final decoding output, restored geometry information (Geometry) and restored attribute information (Attributes) are output.

[0601] Figure 37 shows an encoding method according to embodiments.

[0602] The method according to the embodiments may include a step of encoding geometry data of point cloud data (S3700); and / or a step of encoding attribute data of point cloud data (S3710);

[0603] Figure 38 shows a decryption method according to embodiments.

[0604] The method according to the embodiments may include a step of decoding geometry data of point cloud data in a bitstream (S3810); and / or a step of decoding attribute data of point cloud data (S3810);

[0605] The methods of Fig. 37 and Fig. 38 are described in detail as follows.

[0606] In relation to the decoder (attribute) and ALF operation of FIG. 36 and FIG. 27 (attr_adaptive_loop_filter_enabled), the step of decoding attribute data (S3810) may include: applying a loop filter to the attribute data based on information indicating whether a loop filter in the bitstream is applied to the attribute data.

[0607] With respect to Fig. 27 (attr_alf_component_type) and Fig. 28 (attribute_alf_coeffs), the bitstream further includes information indicating a component type of attribute data to which a loop filter is applied, and the component type to which the loop filter is applied may include at least one of luma and chroma. In addition, the bitstream includes an attribute data unit, and the attribute data unit may include coefficients for the loop filter based on information indicating whether the attribute data is applied.

[0608] With respect to Fig. 29 (attr_alf_window_size, attr_alf_qstep_log2, attr_alf_geom_shift), the bitstream includes an attribute data unit header, and the attribute data unit header may include at least one of a window size for a loop filter, a coefficient quantization step for a loop filter, or a geometry coordinate shift value for a loop filter, based on information indicating whether the attribute data is applied.

[0609] With respect to Fig. 32 (ALF decoding), the step of decoding attribute data (S3810) may further include: obtaining information related to a loop filter in a bitstream; and decoding coefficients related to the loop filter.

[0610] With respect to Fig. 33 (Attribute data unit ALF coefficients syntax), Fig. 11 (cube), and Fig. 34 (NumALFTaps, alf_coeff_abs, alf_coeff_sign), coefficients for the loop filter are obtained based on a cube that divides point cloud data, and coefficients for the loop filter can be derived based on at least one of taps, absolute values, or signs for the loop filter.

[0611] With respect to Fig. 11 (partition), Fig. 13-14 (window size for determining the neighborhood), and Fig. 18-20 (distribution of neighboring voxels for corresponding filter taps), point cloud data is partitioned based on a cube having a specific size, a window size for determining neighboring voxels for predicting point cloud data is derived based on voxel density, and neighboring voxels for filter taps of a loop filter for attribute data can be derived based on at least one of distance or direction.

[0612] A device for performing the method of FIG. 38 includes a memory; and at least one processor connected to the memory; wherein the at least one processor can be configured to: decode geometry data of point cloud data in a bitstream; and decode attribute data of the point cloud data.

[0613] The step (S3710) of encoding attribute data of FIG. 37 may include: applying a loop filter to the attribute data based on information indicating whether a loop filter in the bitstream is applied to the attribute data.

[0614] A device for performing the method of FIG. 37 includes a memory; and at least one processor connected to the memory; wherein the at least one processor can be configured to: encode geometry data of point cloud data; and encode attribute data of the point cloud data.

[0615] The embodiments further include a computer-readable storage medium storing a bitstream generated by the method according to FIG. 37.

[0616] Embodiments further include a method comprising: obtaining a bitstream for point cloud data, the bitstream being generated based on a step of encoding geometry data of the point cloud data; and a step of encoding attribute data of the point cloud data; and transmitting data including the bitstream.

[0617] Hereinafter, the process and effect of attribute coding and decoding of a G-PCC file using geometric transformation prior to a region-based adaptive loop filter are described with reference to data examples of FIGS. 39 to 42.

[0618] Figure 39 shows an example in which Egyptian_mask_vox12 and cognataQC1 have the same scale according to embodiments.

[0619] The method according to the embodiments may further include a geometric transformation before filtration step.

[0620] The method according to the embodiments can encode and decode attribute data using Region based ALF for Reflectance and scant content.

[0621] Using geometric transformations before filtering can significantly improve data quality and image processing.

[0622] Geometric transformations, such as translation, rotation, and scaling, can be used to align images or data before filtering. This allows filters to operate more effectively by making the data more uniform. Transformations can also help improve filtering accuracy by increasing the consistency between the filter and the data structure. In some cases, data may be represented in a form that is inconvenient for filtering. Geometric transformations can transform the data into a more suitable format. Applying geometric transformations can also reduce the computational cost of filtering. For example, reducing the size of an image before filtering reduces the amount of data to be processed, thereby improving processing speed. Overall, using geometric transformations before filtering improves data processing quality, enhances filtering accuracy and efficiency, and reduces computational costs.

[0623] The method according to the embodiments may further include a scaling before filtration step.

[0624] Point clouds are easier to compare when they are at the same scale. This allows you to identify differences and similarities between the two datasets. The criteria for proximity between voxels are different, so the ALF design must also be different. To reduce the variability in how filters are applied, it is recommended to use a set of transformations that alter the geometry. For cognataQC1, the special coordinates are scaled by 64 to align the geometric features, allowing filters with the same settings to be used on both clouds.

[0625] Figure 40 shows an example in which Egyptian_mask_vox12 and cognataQC1 have the same scale.

[0626] The method according to the embodiments may further include a translation before filtration step.

[0627] Applying geometric transformations during data preparation makes the ALF model more robust to geometric characteristics. This increases the occupancy of the region (cube) and reduces overhead.

[0628] Figure 41 shows examples of unfilled blocks and blocks after translation (the number of unprocessed voxels after translation reduces the number of unfilled blocks) according to embodiments.

[0629] Figure 42 shows examples of displayed original blocks and unfilled blocks after translation according to embodiments.

[0630] The method according to the embodiments may further include an affine transform step.

[0631] Applying affine transformation pre-filtering has clear advantages.

[0632] 1. Move, scale, rotate, and translate are basic transformations that can be combined and calculated.

[0633] 2. The inverse transformation of an affine transformation is also an affine transformation.

[0634] 3. Relative proportions and spatial relationships are maintained.

[0635] 4. The spherical neighborhood transforms into an ellipsoid.

[0636] 5. Adjust the parameters of the preprocessed affine transformation to improve the population density of the region and reduce the number of points not processed by the filter.

[0637] 6. This allows the filter design to remain unchanged even for clouds with significantly different geometries.

[0638] Geometric transformations are used in data preparation to enhance the robustness of the ALF model to geometric characteristics. This can increase the occupancy of the region (cube) and reduce overhead.

[0639] By applying adaptive filtering, such as methods such as those in Figures 37 and 38, the visual quality of reconstructed images or videos is significantly improved. Filtering parameters are adjusted based on local characteristics of the content, such as texture, edges, noise levels, and geometry, to more effectively remove artifacts. This can also reduce bitstream size. ALF improves the efficiency of inter-frame prediction by enhancing the quality of reference frames. This reduces the number of bits required to encode inter-frame differences, improving compression efficiency. ALF can be selectively applied to multiple regions of an image or video frame. This allows for targeted enhancements to be implemented where they are most needed without unnecessarily increasing computational complexity or bitrate. Both encoders and decoders implement ALF in a standardized manner, ensuring compatibility and consistent quality across various devices and platforms. ALF can be extended to perform cross-component filtering. By improving the quality of reconstructed point clouds, ALF helps reduce the overall bitrate required for transmission or storage, making it useful for bandwidth-constrained applications.

[0640] The embodiments have been described in terms of methods and / or devices, and the descriptions of methods and devices may be applied complementarily.

[0641] For the convenience of explanation, each drawing has been described separately, but it is also possible to design a new embodiment by combining the embodiments described in each drawing. In addition, designing a computer-readable recording medium having a program recorded thereon for executing the previously described embodiments, as needed by a person skilled in the art, also falls within the scope of the embodiments. The devices and methods according to the embodiments are not limited to the configurations and methods of the embodiments described above, but the embodiments may be configured by selectively combining all or part of the embodiments so that various modifications can be made. Although preferred embodiments of the embodiments have been illustrated and described, the embodiments are not limited to the specific embodiments described above, and various modifications can be made by a person skilled in the art to which the present invention pertains without departing from the gist of the embodiments claimed in the claims, and such modifications should not be understood individually from the technical idea or prospect of the embodiments.

[0642] The various components of the devices of the embodiments may be implemented by hardware, software, firmware, or a combination thereof. The various components of the embodiments may be implemented by a single chip, for example, a single hardware circuit. According to embodiments, the components according to the embodiments may be implemented by separate chips. According to embodiments, at least one of the components of the devices of the embodiments may be configured with one or more processors capable of executing one or more programs, and the one or more programs may perform, or include instructions for performing, one or more of the operations / methods according to the embodiments. The executable instructions for performing the methods / operations of the devices of the embodiments may be stored in non-transitory CRMs or other computer program products configured to be executed by one or more processors, or may be stored in temporary CRMs or other computer program products configured to be executed by one or more processors. In addition, the memory according to the embodiments may be used as a concept including not only volatile memory (e.g., RAM, etc.), but also non-volatile memory, flash memory, PROM, etc. Additionally, it may be implemented in the form of a carrier wave, such as transmission via the Internet. Furthermore, the processor-readable recording medium may be distributed across network-connected computer systems, allowing the processor-readable code to be stored and executed in a distributed manner.

[0643] In this document, " / " and ",” are interpreted as "and / or". For example, "A / B" is interpreted as "A and / or B", and "A, B" is interpreted as "A and / or B". Additionally, "A / B / C" means "at least one of A, B, and / or C". Also, "A, B, C" means "at least one of A, B, and / or C". Additionally, "or" in this document is interpreted as "and / or". For example, "A or B" can mean 1) "A" only, 2) "B" only, or 3) "A and B". In other words, "or" in this document can mean "additionally or alternatively".

[0644] Terms such as "first," "second," etc. may be used to describe various components of the embodiments. However, the various components according to the embodiments should not be interpreted as limited by these terms. These terms are merely used to distinguish one component from another. For example, a first user input signal may be referred to as a "second user input signal." Similarly, a second user input signal may be referred to as a "first user input signal." The use of these terms should be interpreted as not departing from the scope of the various embodiments. Although "first user input signal" and "second user input signal" are both user input signals, they do not mean the same user input signals unless the context clearly indicates otherwise.

[0645] The terminology used to describe the embodiments is for the purpose of describing particular embodiments and is not intended to be limiting of the embodiments. As used in the description of the embodiments and in the claims, the singular is intended to include the plural unless the context clearly dictates otherwise. The expressions “and / or” are used to mean all possible combinations of terms. The expression “includes” describes the presence of features, numbers, steps, elements, and / or components, but does not mean that additional features, numbers, steps, elements, and / or components are not included. Conditional expressions such as “if” or “when” used to describe the embodiments are not intended to be limited to only optional cases. When a specific condition is satisfied, a related action is performed in response to a specific condition, or a related definition is intended to be interpreted.

[0646] Additionally, the operations according to the embodiments described in this document may be performed by a transceiver device including a memory and / or a processor according to the embodiments. The memory may store programs for processing / controlling the operations according to the embodiments, and the processor may control various operations described in this document. The processor may be referred to as a controller, etc. The operations according to the embodiments may be performed by firmware, software, and / or a combination thereof, and the firmware, software, and / or a combination thereof may be stored in the processor or in the memory.

[0647] Meanwhile, the operations according to the embodiments described above may be performed by a transmitting device and / or a receiving device according to the embodiments. The transmitting / receiving device may include a transmitting / receiving unit for transmitting and receiving media data, a memory for storing instructions (program code, algorithm, flowchart, and / or data) for a process according to the embodiments, and a processor for controlling the operations of the transmitting / receiving device.

[0648] The processor may be referred to as a controller or the like, and may correspond to, for example, hardware, software, and / or a combination thereof. The operations according to the above-described embodiments may be performed by the processor. Furthermore, the processor may be implemented as an encoder / decoder or the like for the operations of the above-described embodiments.

[0649] As described above, the relevant contents have been described in the best form for carrying out the embodiments.

[0650] As described above, the embodiments may be applied in whole or in part to a point cloud data transmission and reception device and system.

[0651] Those skilled in the art may make various changes or modifications to the embodiments within the scope of the embodiments.

[0652] Embodiments may include modifications / changes, which do not depart from the scope of the claims and their equivalents.< / int> < / int>

Claims

A step of decoding geometry data of point cloud data in a bitstream; and A step of decoding attribute data of the above point cloud data; comprising; How to decrypt. In the first paragraph, The steps to decode the above attribute data are: applying the loop filter to the attribute data based on information indicating whether the loop filter in the bitstream is applied to the attribute data; How to decrypt. In the second paragraph, The above bitstream further includes information indicating the component type of the attribute data to which the loop filter is applied, The component type to which the loop filter is applied includes at least one of luma and chroma. The above bitstream includes an attribute data unit, The above attribute data unit includes coefficients for the loop filter based on information indicating whether the attribute data is applied to the above attribute data. How to decrypt. In the second paragraph, The above bitstream includes an attribute data unit header, The attribute data unit header includes at least one of a window size for the loop filter, a coefficient quantization step for the loop filter, or a geometry coordinate shift value for the loop filter, based on information indicating whether the attribute data is applied to the attribute data. How to decrypt. In the second paragraph, The steps to decode the above attribute data are: Obtaining information related to the loop filter in the above bitstream; and further comprising decoding coefficients for the above loop filter; How to decrypt. In the third paragraph, The coefficients for the above loop filter are obtained based on a cube that divides the point cloud data, wherein coefficients for the loop filter are derived based on at least one of taps, absolute values, or signs for the loop filter; How to decrypt. In the first paragraph, The above point cloud data is partitioned based on cubes having a specific size, The window size for determining neighboring voxels for predicting the above point cloud data is derived based on the voxel density, Neighboring voxels for filter taps of a loop filter for the above attribute data are derived based on at least one of distance or direction, How to decrypt. In the second paragraph, the method: Before applying the above loop filter, a step of applying a geometric transformation to the point cloud data is further included, Based on the above geometric transformation, the point cloud data is transformed based on at least one of translation, rotation, or scale. How to decrypt. memory; and At least one processor connected to the memory; wherein the at least one processor comprises: Decoding geometry data of point cloud data in the bitstream; and Decoding attribute data of the above point cloud data; configured to do so; Decryption device. A step of encoding geometry data of point cloud data; and A step of encoding attribute data of the above point cloud data; comprising; Encoding method. In paragraph 9, The steps for encoding the above attribute data are: applying the loop filter to the attribute data based on information indicating whether the loop filter in the bitstream is applied to the attribute data; Encoding method. In Article 10, The above geometry data and the above attribute data are included in the bitstream, The above bitstream further includes information indicating the component type of the attribute data to which the loop filter is applied, The component type to which the loop filter is applied includes at least one of luma and chroma. The above bitstream includes an attribute data unit, The above attribute data unit includes coefficients for the loop filter based on information indicating whether the attribute data is applied to the above attribute data. Encoding method. memory; and At least one processor connected to the memory; wherein the at least one processor comprises: Encoding the geometry data of point cloud data; and Encoding attribute data of the above point cloud data; configured to do so; Encoding device. A computer-readable storage medium storing a bitstream generated by the method according to Article 10. Step of obtaining a bitstream for point cloud data, The bitstream is generated based on the steps of encoding geometry data of the point cloud data; and encoding attribute data of the point cloud data; and A method comprising the step of transmitting data including the bitstream.

Images