3D digital human and scene processing method and device, equipment, medium and product

Abstract

The invention provides a 3D digital human and scene processing method, device and equipment, a medium and a product, and the method comprises the steps: obtaining 3D data based on 3D Gaussian sputtering 3DGS; and performing representation generation on the 3D data, specifically, processing the 3D data to obtain a reference point position and a reference point attribute of each reference point, and performing feature transformation based on the reference point position and the reference point attribute in combination with a Gaussian ball position correspondingly generated by the reference point, obtaining an attribute context feature of each datum point; quantizing the datum point attribute of the datum point by using the attribute context feature; and performing entropy coding on the quantized reference point attribute according to the attribute context feature to obtain a coding result.

Description

Technical Field

[0001] The present disclosure relates to the technical field of video transmission, and in particular to a method, device, equipment, medium and product for processing 3D digital humans and scenes. Background Art

[0002] Digital humans refer to virtual characters with human appearance or behavior created using computer technology and artificial intelligence technology. These virtual characters have a wide range of application scenarios, such as corporate publicity, education and training, virtual actors and hosts, intelligent customer service, virtual exhibitions and travel guides, IP image customization, virtual meetings, etc. Recently, 3D digital humans have become a research hotspot. They have the advantages of high realism, strong flexibility, and high cost-effectiveness, and are expected to be widely used in entertainment, education, business, science and other fields.

[0003] In the related art, 3D digital humans can be characterized by 3D Gaussian Splatting (3DGS). However, this characterization method requires a large number of 3D Gaussian functions to well represent large-scale scenes. As a result, each point and its attributes are stored independently, so the characterization method using 3DGS will inevitably bring a lot of storage costs. For example, 3DGS on a typical 360-degree scene requires millions of Gaussian spheres, and its storage space exceeds 1GB. Therefore, the problem of compressing digital humans based on 3DGS characterization is converted into the problem of efficiently compressing Gaussian spheres and their related attributes. Summary of the invention

[0004] The embodiments of the present disclosure at least provide a method, device, equipment, medium and product for processing 3D digital humans and scenes.

[0005] In a first aspect, the embodiments of the present disclosure provide a method for processing a 3D digital human and a scene, including:

[0006] Acquire 3D data based on 3D Gaussian sputtering (3DGS);

[0007] Characterizing and generating the 3D data, specifically comprising: processing the 3D data to obtain a reference point position and a reference point attribute of each reference point, and performing feature transformation based on the reference point position and the reference point attribute in combination with a Gaussian sphere position generated corresponding to the reference point to obtain an attribute context feature of each reference point;

[0008] quantifying a reference point attribute of the reference point using the attribute context feature;

[0009] The quantized reference point attributes are entropy encoded according to the attribute context feature to obtain an encoding result.

[0010] In an optional implementation manner, the processing of the 3D data to obtain a reference point position and a reference point attribute of each reference point includes:

[0011] The 3D data is processed by generating reference points and predicting attributes to obtain the reference point position and reference point attributes of each reference point.

[0012] In an optional implementation manner, the performing of reference point generation and attribute prediction processing on the 3D data to obtain the reference point position and reference point attribute of each reference point includes:

[0013] Converting the 3D data into voxel grid data;

[0014] Based on the position of each voxel center point in the voxel grid data, the reference point position information is determined, and the reference point attribute of the reference point corresponding to the reference point position is determined.

[0015] In an optional implementation, the feature transformation is performed based on the reference point position and the reference point attribute in combination with the Gaussian sphere position generated corresponding to the reference point to obtain the attribute context feature of each reference point, including:

[0016] Based on the reference point position and the reference point attribute, determining the positions of multiple Gaussian spheres corresponding to the reference point;

[0017] Based on each of the Gaussian sphere positions, the reference point positions, and reference point features, attribute context features of the reference point are determined.

[0018] In an optional implementation manner, determining the positions of multiple Gaussian spheres corresponding to the reference point based on the reference point position and the reference point attribute includes:

[0019] Calculating a product between a learnable offset in the fiducial point attribute and a scaling factor in the fiducial point attribute;

[0020] The product and the reference point position are summed to obtain the position of each Gaussian sphere.

[0021] In an optional implementation, determining the attribute context feature of the reference point based on each of the Gaussian sphere positions, the reference point positions, and the reference point features includes:

[0022] According to the position of each Gaussian sphere, determining a preset number of target Gaussian spheres closest to the reference point from among the multiple Gaussian spheres;

[0023] According to the Gaussian sphere position of the target Gaussian sphere and the reference point position and reference point features, attribute context features of the reference point are determined.

[0024] In an optional implementation manner, determining the attribute context feature of the reference point according to the Gaussian sphere position of the target Gaussian sphere and the reference point position and reference point feature includes:

[0025] Determine the distance between the target Gaussian sphere and the reference point according to the Gaussian sphere position of the target Gaussian sphere, and obtain a plurality of first distances;

[0026] Performing a sum operation on the reciprocals of the first distances to obtain a sum result;

[0027] The product of the summation result and the reference point feature is calculated to obtain the attribute context feature of the reference point.

[0028] In an optional implementation manner, the quantifying the reference point attribute of the reference point by using the attribute context feature includes:

[0029] Determining a target quantization step size of the reference point attribute using the attribute context feature;

[0030] The reference point attributes are quantized using the target quantization step size to obtain quantized reference point attributes.

[0031] In an optional implementation manner, determining a target quantization step size of the reference point attribute includes:

[0032] Processing the attribute context feature by a first multi-layer perceptron to obtain a quantization parameter;

[0033] The predefined quantization step size is adjusted by the quantization parameter to obtain the target quantization step size.

[0034] In an optional implementation manner, performing entropy coding on the quantized reference point attribute according to the attribute context feature to obtain a coding result includes:

[0035] Determining the probability distribution of the reference point attribute according to the attribute context feature and the quantized reference point attribute;

[0036] The reference point attributes are entropy encoded using the probability distribution to obtain an encoding result.

[0037] In an optional implementation manner, determining the probability distribution of the reference point attribute according to the attribute context feature and the quantized reference point attribute includes:

[0038] Processing the attribute context feature according to a second multi-layer perceptron to obtain Gaussian distribution parameters of each sub-attribute in the reference point attribute;

[0039] The probability distribution of the reference point attribute is calculated according to the Gaussian distribution parameter, the quantized reference point attribute and the target quantization step size of the reference point attribute.

[0040] In a second aspect, the present disclosure also provides a 3D digital human and scene processing device, including:

[0041] An acquisition unit, used for acquiring 3D data based on three-dimensional Gaussian sputtering 3DGS;

[0042] A representation generation unit is used to generate a representation for the 3D data, specifically comprising: processing the 3D data to obtain a reference point position and a reference point attribute of each reference point, and performing feature transformation based on the reference point position and the reference point attribute in combination with the Gaussian sphere position generated corresponding to the reference point to obtain an attribute context feature of each reference point;

[0043] a quantization unit, configured to quantify a reference point attribute of the reference point using the attribute context feature;

[0044] The entropy coding unit is used to perform entropy coding on the quantized reference point attribute according to the attribute context feature to obtain a coding result.

[0045] In a third aspect, an embodiment of the present disclosure further provides an electronic device, comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, the steps of the above-mentioned first aspect, or any possible implementation of the first aspect are performed.

[0046] In a fourth aspect, an embodiment of the present disclosure further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned first aspect or any possible implementation of the first aspect are executed.

[0047] In a fifth aspect, an embodiment of the present disclosure further provides a computer program product, which is stored in a storage medium and is executed by at least one processor to perform the steps of the above-mentioned first aspect, or any possible implementation of the first aspect.

[0048] The present disclosure provides a method, device, equipment, medium and product for processing 3D digital human and scene. In the embodiment of the present disclosure, after obtaining 3D data based on three-dimensional Gaussian sputtering 3DGS, the 3D data is processed to obtain the reference point position and reference point attribute of each reference point, and based on the reference point position and the reference point attribute, the feature transformation is performed in combination with the Gaussian sphere position generated corresponding to the reference point to obtain the attribute context feature of each reference point; next, the reference point attribute of the reference point is quantized using the attribute context feature; finally, the quantized reference point attribute is entropy encoded according to the attribute context feature to obtain the encoding result.

[0049] In the above implementation, by processing 3D data to obtain the reference point attributes and reference point positions, and determining the Gaussian sphere position corresponding to the reference point, there is no need to transmit all Gaussian sphere information, thereby reducing the amount of data transmitted by encoding; by entropy encoding the quantized reference point attributes through attribute context features, the efficiency of the reference point attribute entropy encoding link can be improved.

[0050] In order to make the above-mentioned objectives, features and advantages of the present disclosure more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following is a brief introduction to the drawings required for use in the embodiments. The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present disclosure and are used together with the specification to illustrate the technical solutions of the present disclosure. It should be understood that the following drawings only illustrate certain embodiments of the present disclosure and should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can also be obtained based on these drawings without creative work.

[0052] Figure 1 A flowchart of a method for processing a 3D digital human and scene provided by an embodiment of the present disclosure is shown;

[0053] Figure 2 A system flow chart of a method for processing a 3D digital human and scene provided by an embodiment of the present disclosure is shown;

[0054] Figure 3 A schematic diagram of a 3D digital human and scene processing device provided by an embodiment of the present disclosure is shown;

[0055] Figure 4 A schematic diagram of an electronic device provided by an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical scheme and advantages of the embodiments of the present disclosure clearer, the technical scheme in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all of the embodiments. The components of the embodiments of the present disclosure generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present disclosure provided in the drawings is not intended to limit the scope of the present disclosure for protection, but merely represents the selected embodiments of the present disclosure. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present disclosure.

[0057] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0058] The term "and / or" herein only describes an association relationship, indicating that three relationships may exist. For example, A and / or B may represent the following three situations: A exists alone, A and B exist at the same time, and B exists alone. In addition, the term "at least one" herein represents any combination of at least two of any one or more of a plurality of. For example, including at least one of A, B, and C may represent including any one or more elements selected from the set consisting of A, B, and C.

[0059] From the above description, we can see that the 3DGS representation method requires a large number of 3D Gaussian functions to well represent large-scale scenes. As a result, each point and its attributes are stored independently, so the use of the 3DGS representation method will inevitably bring a lot of storage costs. For example, a typical 360-degree scene requires millions of Gaussian spheres, and its storage space exceeds 1GB. Therefore, the problem of compressing digital humans based on 3DGS representation is converted into the problem of efficiently compressing Gaussian spheres and their related attributes.

[0060] Based on the above research, the present disclosure provides a method, device, equipment, medium and product for processing 3D digital people and scenes. In the embodiment of the present disclosure, after obtaining 3D data based on three-dimensional Gaussian sputtering 3DGS, the 3D data is characterized and generated, specifically including: processing the 3D data to obtain the reference point position and reference point attributes of each reference point, and based on the reference point position and the reference point attributes, combining the Gaussian sphere position generated corresponding to the reference point, perform feature transformation to obtain the attribute context feature of each reference point; next, quantize the reference point attributes of the reference point using the attribute context feature; finally, entropy code the quantized reference point attributes according to the attribute context feature to obtain the coding result.

[0061] In the above implementation, by processing 3D data to obtain the reference point attributes and reference point positions, and determining the Gaussian sphere position corresponding to the reference point, there is no need to transmit all Gaussian sphere information, thereby reducing the amount of data transmitted by encoding; by entropy encoding the quantized reference point attributes through attribute context features, the efficiency of the reference point attribute entropy encoding link can be improved.

[0062] To facilitate understanding of this embodiment, a method for processing a 3D digital human and a scene disclosed in the embodiment of the present disclosure is first introduced in detail. The execution subject of the method for processing a 3D digital human and a scene provided in the embodiment of the present disclosure is generally an electronic device with certain computing capabilities. In some possible implementations, the method for processing a 3D digital human and a scene can be implemented by a processor calling a computer-readable instruction stored in a memory. The embodiment of the present disclosure can be applied to the data compression process of a 3D digital human and related scenes (for example, a scene). The related scenes include related three-dimensional scenes such as 3D objects and 3D spaces.

[0063] See also Figure 1 FIG. 1 is a flowchart of a method for processing a 3D digital human and a scene provided by an embodiment of the present disclosure. The method includes steps S101 to S104, wherein:

[0064] S101: Acquire 3D data based on three-dimensional Gaussian sputtering (3DGS).

[0065] Here, 3DGS uses a set of anisotropic 3D Gaussian functions to represent the scene, which have the differentiable properties of volume representation and can be efficiently rendered by tile-based rasterization.

[0066] Three-dimensional Gaussian sputtering is an emerging way to represent 3D digital humans. 3DGS models the three-dimensional world as a mixture of multiple 3D Gaussians. In theory, a mixture of multiple Gaussian distributions can fit any distribution. 3D Gaussians model the shapes of scene objects. In terms of object color modeling, 3DGS uses spherical harmonics, which can express the value size of a sphere in various directions. This value represents the RGB value of the corresponding position. In summary, the shape is explicitly represented by 3D Gaussians, and the color is represented by spherical harmonics. In this way, through the projection relationship, the three-dimensional scene can be rendered to visualize the image at any perspective. 3DGS can achieve high-quality real-time rendering with a resolution of 1080p, and has brought revolutionary improvements in the control and editability of the scene.

[0067] S102: Characterizing and generating the 3D data, specifically comprising: processing the 3D data to obtain a reference point position and reference point attributes of each reference point, and performing feature transformation based on the reference point position and the reference point attributes in combination with a Gaussian sphere position generated corresponding to the reference point to obtain an attribute context feature of each of the reference points.

[0068] In the disclosed embodiment, 3D data can be processed by a characterization generation module to obtain the reference point position and reference point attributes of each reference point, and based on the reference point position and reference point attributes, feature transformation is performed in combination with the Gaussian sphere position generated corresponding to the reference point to obtain the attribute context feature of each reference point.

[0069] Here, the sparse point cloud collected by the 3D digital human signal can be used as the input of the representation generation module. Through the representation generation module, sparse voxel grid data can be generated based on the sparse point cloud, so as to determine the reference point according to the center point of each voxel, and determine the reference point position and reference point attribute of the reference point; wherein, the reference point attribute Contains the following: Benchmark point features Scaling Factor and k learnable offsets Among them, D v is the vector f v Dimension.

[0070] For each reference point, at least one Gaussian sphere can be generated. In the disclosed embodiment, the distance between the reference point and the corresponding Gaussian sphere and the reference point feature of the reference point can be weighted to obtain the attribute context feature. The attribute context feature can be used to predict the probability distribution of the reference point attribute to assist the subsequent entropy coding probability distribution, thereby achieving a better coding effect.

[0071] S103: quantifying the reference point attributes of the reference point using the attribute context feature.

[0072] In order to facilitate entropy coding, the reference point attributes need to be The value of is quantized into a finite set. The technical solution disclosed in the present invention does not limit the quantization method. For example, a rounding quantization algorithm, an adaptive quantization algorithm, and other methods may be used. The quantization algorithm is not specifically limited here, and is subject to implementation.

[0073] S104: performing entropy coding on the quantized reference point attributes according to the attribute context features to obtain a coding result.

[0074] In the disclosed embodiment, the probability distribution of the reference point attributes may be determined according to the attribute context feature, and then the reference point attributes may be entropy encoded according to the probability distribution to obtain an encoding result.

[0075] In the above implementation, by processing 3D data to obtain the reference point attributes and reference point positions, and determining the Gaussian sphere position corresponding to the reference point, there is no need to transmit all Gaussian sphere information, thereby reducing the amount of data transmitted by encoding; by entropy encoding the quantized reference point attributes through attribute context features, the efficiency of the reference point attribute entropy encoding link can be improved.

[0076] The above process will be described in detail below in conjunction with specific implementation methods.

[0077] It can be seen from the above description that in the embodiment of the present disclosure, after obtaining 3D data based on three-dimensional Gaussian sputtering 3DGS, the 3D data can be processed by a characterization generation module to obtain the reference point position and reference point attributes of each reference point. The characterization generation module includes the following modules: a reference point generation and attribute prediction module and a feature transformation module.

[0078] On this basis, the above steps process the 3D data to obtain the reference point position and reference point attributes of each reference point, which specifically includes the following steps:

[0079] The 3D data is processed by generating reference points and predicting attributes to obtain the reference point position and reference point attributes of each reference point.

[0080] In the embodiment of the present disclosure, after the 3D data is acquired, the 3D data may be processed, and the processing result may be processed to predict the reference point generation attributes to obtain the reference point position and reference point attributes of each reference point.

[0081] In the embodiment of the present disclosure, the above steps perform reference point generation and attribute prediction processing on the 3D data to obtain the reference point position and reference point attribute of each reference point, specifically including:

[0082] Step S11: converting the 3D data into voxel grid data;

[0083] Step S12: determining the reference point position information based on the position of each voxel center point in the voxel grid data, and determining the reference point attribute of the reference point corresponding to the reference point position.

[0084] In the embodiment of the present disclosure, a structure from motion (SFM) method is used to generate voxel grid data of 3D data; wherein the center point of each voxel can be used as a reference point v (anchor), that is, the position of each center point determines the reference point position of the reference point; wherein each reference point has a reference point feature Scaling Factor and k learnable offsets These attributes.

[0085] After obtaining the reference point position and reference point attributes of each reference point, it is also necessary to determine the positions of multiple Gaussian spheres corresponding to the reference point and the attribute characteristics of the Gaussian spheres.

[0086] Specifically, for position x c The observation point (camera) and position x at v The relative distance and viewing direction of the reference points can be calculated using the following formula:

[0087] δ vc =||x v -x c2 ||2, The relative distance and viewing direction are used to determine the attribute characteristics of the Gaussian sphere corresponding to the reference point, and the attribute characteristics of the Gaussian sphere are used to render a three-dimensional scene in the rendering process of the 3D digital human to visualize the image of the 3D digital human at any viewing angle.

[0088] In the above-mentioned benchmark point generation and attribute prediction process, the center of each voxel in the above-mentioned voxel grid data is used as the benchmark point; wherein each benchmark point is connected to a set of Gaussian functions with learnable offsets, and its attribute characteristics are further dynamically predicted based on the benchmark point characteristics and viewing angle position.

[0089] In an optional implementation, the above steps are based on the reference point position and the reference point attribute, combined with the Gaussian sphere position generated corresponding to the reference point, to perform feature transformation to obtain the attribute context feature of each reference point, specifically including the following steps:

[0090] Step S21: determining the positions of multiple Gaussian spheres corresponding to the reference points based on the reference point positions and the reference point attributes;

[0091] Step S22: Determine the attribute context feature of the reference point based on each of the Gaussian sphere positions, the reference point positions and reference point features.

[0092] In the embodiment of the present disclosure, after the reference point is determined, the 3DGS corresponding to the scene (i.e., the 3D Gaussian sphere corresponding to the reference point) can be generated from the reference point, and the attribute characteristics of each 3D Gaussian sphere can be determined, through which the parameterization of the Gaussian function can be realized. Among them, the attribute characteristics of the 3D Gaussian sphere specifically include the following: opacity Covariance-related quaternions Zoom And color

[0093] Here, the reference point position and some of the attributes in the reference point attributes can be combined to determine the Gaussian sphere position corresponding to the reference point. Through each individual multi-layer perceptron (MLP), the reference point feature f can be directly obtained. v , the relative viewing distance δ between the camera and the reference point vc and viewing direction The attribute features of k 3D Gaussian balls are obtained by decoding. The MLP used to calculate each attribute feature can be recorded as F α , F c , F q and F s , which are used to decode opacity, color, quaternion and scaling features respectively.

[0094] For example, the opacity value of a 3D Gaussian sphere generated from a reference point can be calculated using the following formula:

[0095] Corresponding to other attribute features of the 3D Gaussian ball, such as color {c i}、quaternion{q i} and scaling features {s i} are derived in a similar way, that is, the formula can be expressed as: and

[0096] After the position of the Gaussian sphere is determined in the above manner, the distance between the reference point and the corresponding Gaussian sphere and the reference point feature of the reference point may be weighted to obtain the attribute context feature.

[0097] In the embodiment of the present disclosure, the above steps determine the positions of multiple Gaussian spheres corresponding to the reference points based on the reference point positions and the reference point attributes, specifically including:

[0098] First, the product between the learnable offset in the reference point attribute and the scaling factor in the reference point attribute is calculated;

[0099] Secondly, the product and the reference point position are summed to obtain the position of each Gaussian sphere.

[0100] It can be seen from the above description that the technical solution disclosed in the present invention generates k 3D Gaussian spheres for each visible reference point in the viewing cone and predicts their attribute features.

[0101] Here, we need to combine the reference point position and reference point attributes to determine the k 3D Gaussian spheres corresponding to the reference point. Specifically, given a point at x v The calculation formula of the 3D Gaussian sphere position is as follows:

[0102] in, is the learnable offset (k offsets), l v is the scaling factor relative to this reference point.

[0103] Here, first, the learnable offset O in the reference point attribute is calculated v With scaling factor l v The product between Then, the sum of the product and the reference point position is calculated to obtain the position of each Gaussian ball {μ0, ..., μ k-1}.

[0104] In the embodiment of the present disclosure, the above steps determine the attribute context feature of the reference point based on the position of each Gaussian sphere, the position of the reference point and the reference point feature, specifically including:

[0105] First, according to the position of each Gaussian sphere, a preset number of target Gaussian spheres closest to the reference point are determined from the multiple Gaussian spheres;

[0106] Secondly, the attribute context feature of the reference point is determined according to the Gaussian sphere position of the target Gaussian sphere and the reference point position and reference point feature.

[0107] After the benchmark point generation and attribute prediction are completed, the distance between the Gaussian ball generated corresponding to the benchmark point and the benchmark point can be compared with the benchmark point feature f v After weighting, we get the attribute context feature f h , and use it as context to predict the probability distribution of the benchmark point attributes to assist the subsequent entropy coding probability distribution to achieve better coding effects.

[0108] From the above description, we can know that each reference point is composed of the reference point position and datum point properties The components of the reference point attributes represent the reference point features, scaling factors, and learnable offsets. For each reference point, k Gaussian balls are generated. Here, m Gaussian balls closest to the reference point can be selected as target Gaussian balls, where m = [k / 3]. Then, the attribute context feature f is determined based on the position of the target Gaussian ball, the reference point position, and the reference point features. h Specifically, the distance between the target Gaussian ball and the reference point can be compared with the reference point feature f v After weighting, we get the attribute context feature f h .

[0109] In the embodiment of the present disclosure, the above steps determine the attribute context feature of the reference point according to the Gaussian sphere position of the target Gaussian sphere and the reference point position and reference point feature, and specifically include the following steps:

[0110] First, according to the Gaussian sphere position of the target Gaussian sphere, the distance between the target Gaussian sphere and the reference point is determined to obtain a plurality of first distances;

[0111] Secondly, performing a sum operation on the reciprocals of the first distances to obtain a sum result;

[0112] Finally, the product of the summation result and the reference point feature is calculated to obtain the attribute context feature of the reference point.

[0113] In the embodiment of the present disclosure, the distance from the target Gaussian sphere to the reference point (i.e., the first distance) can be determined according to the position of the target Gaussian sphere and the position of the reference point. The first distances are d1, d2, d3, ..., d m Next, calculate the inverse of each first distance, respectively Then, the reciprocal of the first distance is summed to obtain the sum result Finally, the product between this summation and the reference point feature can be calculated Thus, we get the attribute context feature f h .

[0114] Therefore, the above process can be described by the formula:

[0115] In the technical solution disclosed in the present invention, the attribute context feature is weighted by the reference point position and reference point features, combined with the distance between the generated Gaussian sphere and the reference point, and used as the context to predict the probability distribution of the reference point attribute, which can improve the efficiency of subsequent entropy coding.

[0116] In the embodiment of the present disclosure, after the attribute context feature is obtained by the representation generation module, the attribute context feature can be used to quantify the reference point attribute of the reference point, which specifically includes the following steps:

[0117] Step S31: Determine the target quantization step size of the reference point attribute by using the attribute context feature;

[0118] Step S32: quantizing the reference point attributes using the target quantization step size to obtain quantized reference point attributes.

[0119] Here, we get the attribute context feature f h After that, we need to use the attribute context feature f h Minimize datum point properties The entropy of the reference point attributes is increased, thereby reducing the number of bits when entropy coding is used to encode the reference point attributes.

[0120] To facilitate entropy coding, the reference point attributes can be The value of is quantized into a finite set. The present invention does not limit the quantization method. For example, various methods such as traditional rounding and self-quantization can be used. The following introduces an adaptive quantization algorithm that can be used.

[0121] In the disclosed embodiment, the adaptive quantization algorithm can use MLP to adaptively determine the quantization step size to obtain the target quantization step size. Then, the reference point attribute can be quantized by the target quantization step size to obtain the quantized reference point attribute

[0122] In the embodiment of the present disclosure, the above step of determining the target quantization step size of the reference point attribute specifically includes:

[0123] Processing the attribute context feature by a first multi-layer perceptron to obtain a quantization parameter;

[0124] The predefined quantization step size is adjusted by the quantization parameter to obtain the target quantization step size.

[0125] In the above adaptive quantization method, MLPq (i.e., the first multi-layer perceptron) can be used to estimate the quantization parameters. In the quantization process, the context model MLPq and the attribute context feature f i h Input, predict the quantization parameters The specific formula can be described as r i =MLP q (f i h ); wherein the quantization parameter is used to adjust the predefined quantization step size Q0.

[0126] After obtaining the quantization parameter r i Afterwards, the predefined quantization step size Q0 is adjusted according to the quantization parameter to obtain the target quantization step size. The specific adjustment formula is described as follows: i =Q0×(1+Tanh(r i )).

[0127] After the reference point attributes of the reference point are quantized, entropy coding may be performed on the quantized reference point attributes according to attribute context features to obtain a coding result, which specifically includes the following steps:

[0128] Step S41: determining the probability distribution of the reference point attributes according to the attribute context features and the quantized reference point attributes;

[0129] Step S42: performing entropy coding on the reference point attributes through the probability distribution to obtain a coding result.

[0130] In the embodiment of the present disclosure, after obtaining the target quantization step size q i Afterwards, the reference point attributes may be quantized according to the target quantization step size, thereby obtaining quantized reference point attributes.

[0131] Here, an MLP can be used to model Gaussian parameters, denoted as MLP c , whose structure can be compared with the quantized context model MLP q The context model MLP with the Gaussian parameters is c It can be used to determine the probability distribution of the reference point attributes. Specifically, the context model MLP of the Gaussian parameters c Used to model Gaussian parameters to obtain the probability distribution of benchmark point attributes.

[0132] Next, the reference point attributes can be entropy encoded using the probability distribution to obtain an encoding result. By using the probability distribution to assist the entropy encoding of the quantized reference point attributes, a better encoding effect can be achieved.

[0133] In the embodiment of the present disclosure, the above step of determining the probability distribution of the reference point attribute according to the attribute context feature and the quantized reference point attribute specifically includes:

[0134] First, the attribute context feature is processed according to a second multi-layer perceptron to obtain Gaussian distribution parameters of each sub-attribute in the reference point attribute;

[0135] Secondly, the probability distribution of the reference point attribute is calculated according to the Gaussian distribution parameters, the quantized reference point attribute and the target quantization step size of the reference point attribute.

[0136] Through the above description, the reference point attributes include: reference point characteristics Scaling Factor and k learnable offsets Here, the reference point features, scaling factors and offsets are the sub-attributes described in the above steps. The second multi-layer perceptron is the context model MLP of the Gaussian parameters described above. c .

[0137] Datum point properties All three sub-attributes of show the statistical trend of Gaussian distribution. However, using the same set of Gaussian distribution parameters μ and σ for different sub-attributes may lack accuracy. Therefore, the technical solution of the present disclosure assumes that the reference point attribute The values ​​of each sub-attribute are independent of each other. At this time, it is necessary to determine the Gaussian distribution parameters of each sub-attribute. The Gaussian distribution parameters μ and σ of each sub-attribute are based on the context model MLP of MLP. c From h Estimated.

[0138] After obtaining the Gaussian distribution parameters of each sub-attribute, the probability density function of the reference point can be determined according to the Gaussian distribution parameters. and the cumulative distribution function Next, we can use the probability density function and the cumulative distribution function As well as the quantized benchmark point attribute and the target quantization step of the benchmark point attribute, the probability distribution of the benchmark point attribute is calculated.

[0139] Specifically, for the i-th reference point and its quantized reference point attribute vector The Gaussian distribution parameter of each sub-attribute can be calculated by the following formula: i ,σ i =MLP c (f i h ). Then, using the estimated Can be calculated The probability of is calculated as follows:

[0140]

[0141] From the above description, it can be seen that the technical solution disclosed in the present invention uses the context model MLP of Gaussian parameters c The attribute context features are processed to obtain Gaussian distribution parameters of each reference point attribute; then, the probability distribution of the reference point attribute is obtained through the Gaussian distribution parameters, the quantized attribute context features and the target quantization step size, and then the reference point attribute is entropy encoded according to the probability distribution. This processing method can improve the efficiency of the reference point attribute entropy encoding link.

[0142] The following will be combined Figure 2 Describe the above steps. Figure 2 The system flow chart of the method for processing 3D digital human and scene is shown as follows: Figure 2 As shown, the method mainly includes the following processes: characterization generation, quantization and entropy coding; wherein, characterization generation includes SFM processing, reference point generation and attribute prediction, and feature transformation; the specific implementation process of the above process is described as follows:

[0143] The process of characterization generation:

[0144] First, the motion recovery structure SFM method is used to generate sparse voxel grid data of 3D data, where the center point of each voxel can be regarded as a reference point v (anchor), which has the characteristics of the reference point Scaling Factor and k learnable offsets

[0145] Specifically, for position x c The observation point (camera) and position x at v The relative distance and viewing direction of the reference points can be calculated using the following formula:

[0146] The relative distance and the viewing direction are used to determine the attribute characteristics of the Gaussian sphere corresponding to the reference point, and the attribute characteristics of the Gaussian sphere are the characteristics used in the rendering process of the 3D digital human.

[0147] After the reference point is determined, the 3DGS corresponding to the scene is generated from the reference point (i.e., the 3D Gaussian sphere corresponding to the reference point), and the attribute characteristics of each 3D Gaussian sphere are determined, through which the parameterization of the Gaussian function can be realized. Among them, the attribute characteristics of the 3D Gaussian sphere specifically include the following: opacity Covariance-related quaternions Zoom And color

[0148] Here, the reference point position and reference point attributes can be combined to determine the Gaussian sphere position corresponding to the reference point. Specifically, given the position at x v The calculation formula of the corresponding Gaussian sphere position is as follows:

[0149] Through each individual multi-layer perceptron (MLP), we can directly obtain the reference point features f v , the relative viewing distance δ between the camera and the reference point vc and viewing direction The attribute features of k 3D Gaussian balls are obtained by decoding. The MLP used to calculate each attribute feature can be recorded as F α , F c , F q and F s , which are used to decode opacity, color, quaternion and scaling features respectively.

[0150] For example, the opacity value of a 3D Gaussian sphere generated from a reference point can be calculated using the following formula: Corresponding to other attribute features of the 3D Gaussian ball, such as color {c i}、quaternion{qi} and scaling features {s i} are derived in a similar way and will not be repeated here.

[0151] Next, for each reference point, k Gaussian balls are generated, and m Gaussian balls closest to the reference point are selected as target Gaussian balls, where m = [k / 3], and the first distance between each target Gaussian ball and the reference point is d1, d2, d3, ..., d m The inverse of the first distance and the reference point feature f are combined according to the following formula: v Weighted attribute context feature f h :

[0152]

[0153] The process of quantization and entropy coding:

[0154] After the attribute context feature is obtained, the attribute context feature can be used to minimize the entropy of the reference point attribute, thereby reducing the number of bits when the reference point attribute is encoded using entropy coding.

[0155] To facilitate entropy coding, the reference point attributes can be The value of is quantized to a finite set. For example, various methods such as traditional rounding and self-quantization can be used.

[0156] In the above adaptive quantization method, MLPq can be used to estimate the quantization parameters. In the quantization process, the context model MLPq and the attribute context feature f i h Input, predict the quantization parameters The specific formula can be described as r i =MLP q (f i h ); wherein the quantization parameter is used to adjust the predefined quantization step size Q0. Then, the predefined quantization step size Q0 is adjusted according to the quantization parameter, thereby obtaining the target quantization step size. The specific adjustment formula is described as follows: i =Q0×(1+Tanh(r i )). After obtaining the target quantization step size, the reference point attributes can be quantized according to the target quantization step size to obtain the quantized reference point attributes.

[0157] Next, we can use the context model MLP cThe attribute context features are processed to obtain the Gaussian distribution parameters of each reference point attribute; the probability distribution of the reference point attribute is obtained through the Gaussian distribution parameters, the quantized attribute context features and the target quantization step size, and then the reference point attribute is entropy encoded according to the probability distribution to obtain the encoding result, that is, Figure 2 The specific formula for the compressed code stream shown is as follows:

[0158]

[0159] μ i ,σ i =MLP c (f i h )

[0160] In the disclosed embodiment, entropy coding can be optimized and updated by using a deep learning-based encoder or directly using an entropy function as a loss function. It should be noted that in MLP training, a single loss function or a combination of multiple loss functions can be used, such as an entropy loss function, a rendering and reconstruction loss function, etc.

[0161] From the above description, it can be seen that the technical solution disclosed in the present invention uses the sparse point cloud collected by the 3D digital human signal as the input for representation generation, and adopts the motion recovery structure SFM method to generate sparse voxel grid data of 3D data. In the process of reference point generation and attribute prediction, the center of each voxel in the sparse voxel grid data is used as the reference point. For each reference point, k Gaussian balls can be generated, and the attribute characteristics of each Gaussian ball can be dynamically predicted according to the reference point characteristics and the viewing angle position.

[0162] Then, the generated reference point positions and reference point attributes are input into the feature transformation process for processing. Specifically, the reference point positions and reference point attributes are weighted by the distance between the generated Gaussian sphere and the reference point to obtain the attribute context features. The reference point attributes are quantized through the quantization process to obtain the quantized reference point attributes. Next, the attribute context features output by the feature change process are used as context to predict the probability distribution of the reference point attributes, thereby improving the efficiency of the reference point attribute entropy coding link.

[0163] The technical solution disclosed in the present invention proposes a digital human compression solution based on 3DGS, which converts the digital human data compression problem into Gaussian sphere compression; the overall system uses the sparse point cloud of the 3D data of the 3D digital human as input, and finally outputs the entropy coded code stream of the reference point position of the 3D digital human and its attributes through the above process, thereby reducing the amount of data in network transmission.

[0164] Those skilled in the art will appreciate that, in the above method of specific implementation, the order in which the steps are written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of the steps should be determined by their functions and possible internal logic.

[0165] Based on the same inventive concept, the embodiments of the present disclosure also provide a 3D digital human and scene processing device corresponding to the 3D digital human and scene processing method. Since the principle of solving the problem by the device in the embodiments of the present disclosure is similar to the above-mentioned 3D digital human and scene processing method in the embodiments of the present disclosure, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.

[0166] Reference Figure 3 FIG. 1 is a schematic diagram of a 3D digital human and scene processing device provided by an embodiment of the present disclosure, wherein the device comprises: an acquisition module 10, a representation generation module 20, a quantization module 30 and an entropy coding module 40; wherein,

[0167] An acquisition module 10 is used to acquire 3D data based on three-dimensional Gaussian sputtering 3DGS;

[0168] The representation generation module 20 is used to generate a representation for the 3D data, specifically comprising: processing the 3D data to obtain a reference point position and a reference point attribute of each reference point, and performing feature transformation based on the reference point position and the reference point attribute in combination with the Gaussian sphere position generated corresponding to the reference point to obtain an attribute context feature of each reference point;

[0169] A quantification module 30, configured to quantify the reference point attributes of the reference point using the attribute context features;

[0170] The entropy coding module 40 is used to perform entropy coding on the quantized reference point attributes according to the attribute context features to obtain a coding result.

[0171] In a possible implementation manner, the representation generation module is further used to:

[0172] The 3D data is processed by generating reference points and predicting attributes to obtain the reference point position and reference point attributes of each reference point.

[0173] In a possible implementation, the representation generation module is further used to: convert the 3D data into voxel grid data;

[0174] Based on the position of each voxel center point in the voxel grid data, the reference point position information is determined, and the reference point attribute of the reference point corresponding to the reference point position is determined.

[0175] In a possible implementation manner, the representation generation module is further used to:

[0176] Based on the reference point position and the reference point attribute, determining the positions of multiple Gaussian spheres corresponding to the reference point;

[0177] Based on each of the Gaussian sphere positions, the reference point positions, and reference point features, attribute context features of the reference point are determined.

[0178] In a possible implementation manner, the representation generation module is further used to:

[0179] According to the position of each Gaussian sphere, determining a preset number of target Gaussian spheres closest to the reference point from among the multiple Gaussian spheres;

[0180] According to the Gaussian sphere position of the target Gaussian sphere and the reference point position and reference point features, attribute context features of the reference point are determined.

[0181] In a possible implementation manner, the representation generation module is further used to:

[0182] Determine the distance between the target Gaussian sphere and the reference point according to the Gaussian sphere position of the target Gaussian sphere, and obtain a plurality of first distances;

[0183] Performing a sum operation on the reciprocals of the first distances to obtain a sum result;

[0184] The product of the summation result and the reference point feature is calculated to obtain the attribute context feature of the reference point.

[0185] In a possible implementation manner, the quantization module is further used for:

[0186] Determining a target quantization step size of the reference point attribute using the attribute context feature;

[0187] The reference point attributes are quantized using the target quantization step size to obtain quantized reference point attributes.

[0188] In a possible implementation manner, the quantization module is further used for:

[0189] Processing the attribute context feature by a first multi-layer perceptron to obtain a quantization parameter;

[0190] The predefined quantization step size is adjusted by the quantization parameter to obtain the target quantization step size.

[0191] In a possible implementation manner, the entropy coding module is further used for:

[0192] Determining the probability distribution of the reference point attribute according to the attribute context feature and the quantized reference point attribute;

[0193] The reference point attributes are entropy encoded using the probability distribution to obtain an encoding result.

[0194] In a possible implementation manner, the entropy coding module is further used for:

[0195] Processing the attribute context feature according to a second multi-layer perceptron to obtain Gaussian distribution parameters of each sub-attribute in the reference point attribute;

[0196] The probability distribution of the reference point attribute is calculated according to the Gaussian distribution parameter, the quantized reference point attribute and the target quantization step size of the reference point attribute.

[0197] For descriptions of the processing flow of each module in the device and the interaction flow between each module, reference may be made to the relevant descriptions in the above method embodiment, which will not be described in detail here.

[0198] Corresponds to Figure 1 The present disclosure also provides an electronic device 400, such as Figure 4 FIG. 4 is a schematic diagram of the structure of an electronic device 400 provided in an embodiment of the present disclosure, including:

[0199] Processor 41, memory 42, and bus 43; memory 42 is used to store execution instructions, including internal memory 421 and external memory 422; the internal memory 421 is also called internal memory, which is used to temporarily store the operation data in the processor 41 and the data exchanged with the external memory 422 such as a hard disk. The processor 41 exchanges data with the external memory 422 through the internal memory 421. When the electronic device 400 is running, the processor 41 communicates with the memory 42 through the bus 43, so that the processor 41 executes the following instructions:

[0200] Acquire 3D data based on 3D Gaussian sputtering (3DGS);

[0201] Characterizing and generating the 3D data, specifically comprising: processing the 3D data to obtain a reference point position and a reference point attribute of each reference point, and performing feature transformation based on the reference point position and the reference point attribute in combination with a Gaussian sphere position generated corresponding to the reference point to obtain an attribute context feature of each reference point;

[0202] quantifying a reference point attribute of the reference point using the attribute context feature;

[0203] The quantized reference point attributes are entropy encoded according to the attribute context feature to obtain an encoding result.

[0204] The embodiment of the present disclosure also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method for processing 3D digital humans and scenes described in the above method embodiment are executed. The storage medium can be a volatile or non-volatile computer-readable storage medium.

[0205] The embodiments of the present disclosure also provide a computer program product, which carries a program code. The instructions included in the program code can be used to execute the steps of the 3D digital human and scene processing method described in the above method embodiment. For details, please refer to the above method embodiment, which will not be repeated here.

[0206] The computer program product may be implemented in hardware, software or a combination thereof. In one optional embodiment, the computer program product is implemented as a computer storage medium. In another optional embodiment, the computer program product is implemented as a software product, such as a software development kit (SDK).

[0207] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the system and device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. In the several embodiments provided in the present disclosure, it should be understood that the disclosed system, device and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of the device or unit can be electrical, mechanical or other forms.

[0208] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0209] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0210] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present disclosure, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present disclosure. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0211] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present disclosure, which are used to illustrate the technical solutions of the present disclosure, rather than to limit them. The protection scope of the present disclosure is not limited thereto. Although the present disclosure is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed in the present disclosure, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure, and should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A method for processing 3D digital humans and scenes, characterized in that: include: Acquire 3D data based on 3D Gaussian sputtering (3DGS); Characterizing and generating the 3D data, specifically comprising: processing the 3D data to obtain a reference point position and a reference point attribute of each reference point, and performing feature transformation based on the reference point position and the reference point attribute in combination with a Gaussian sphere position generated corresponding to the reference point to obtain an attribute context feature of each reference point; quantifying a reference point attribute of the reference point using the attribute context feature; The quantized reference point attributes are entropy encoded according to the attribute context feature to obtain an encoding result.

2. The method according to claim 1, characterized in that The processing of the 3D data to obtain the reference point position and reference point attributes of each reference point includes: The 3D data is processed by generating reference points and predicting attributes to obtain the reference point position and reference point attributes of each reference point.

3. The method according to claim 2, characterized in that The step of performing benchmark point generation and attribute prediction processing on the 3D data to obtain a benchmark point position and a benchmark point attribute of each benchmark point includes: Converting the 3D data into voxel grid data; Based on the position of each voxel center point in the voxel grid data, the reference point position information is determined, and the reference point attribute of the reference point corresponding to the reference point position is determined.

4. The method according to claim 1, characterized in that: The feature transformation is performed based on the reference point position and the reference point attribute in combination with the Gaussian sphere position generated corresponding to the reference point to obtain the attribute context feature of each reference point, including: Based on the reference point position and the reference point attribute, determining the positions of multiple Gaussian spheres corresponding to the reference point; Based on each of the Gaussian sphere positions, the reference point positions, and reference point features, attribute context features of the reference point are determined.

5. The method according to claim 4, characterized in that The determining, based on the reference point position and the reference point attribute, positions of multiple Gaussian spheres corresponding to the reference point includes: Calculating a product between a learnable offset in the fiducial point attribute and a scaling factor in the fiducial point attribute; The product and the reference point position are summed to obtain the position of each Gaussian sphere.

6. The method according to claim 4, characterized in that The determining the attribute context feature of the reference point based on each of the Gaussian sphere positions, the reference point position, and the reference point feature comprises: According to the position of each Gaussian sphere, determining a preset number of target Gaussian spheres closest to the reference point from among the multiple Gaussian spheres; According to the Gaussian sphere position of the target Gaussian sphere and the reference point position and reference point features, attribute context features of the reference point are determined.

7. The method according to claim 6, characterized in that The determining the attribute context feature of the reference point according to the Gaussian sphere position of the target Gaussian sphere and the reference point position and reference point feature includes: Determine the distance between the target Gaussian sphere and the reference point according to the Gaussian sphere position of the target Gaussian sphere, and obtain a plurality of first distances; Performing a sum operation on the reciprocals of the first distances to obtain a sum result; The product of the summation result and the reference point feature is calculated to obtain the attribute context feature of the reference point.

8. The method according to claim 1, characterized in that The step of quantifying the reference point attribute of the reference point by using the attribute context feature includes: Determining a target quantization step size of the reference point attribute using the attribute context feature; The reference point attributes are quantized using the target quantization step size to obtain quantized reference point attributes.

9. The method according to claim 8, characterized in that The determining of the target quantization step size of the reference point attribute comprises: Processing the attribute context feature by a first multi-layer perceptron to obtain a quantization parameter; The predefined quantization step size is adjusted by the quantization parameter to obtain the target quantization step size.

10. The method according to claim 1, characterized in that The entropy encoding of the quantized reference point attribute according to the attribute context feature to obtain an encoding result includes: Determining the probability distribution of the reference point attribute according to the attribute context feature and the quantized reference point attribute; The reference point attributes are entropy encoded using the probability distribution to obtain an encoding result.

11. The method according to claim 10, characterized in that Determining the probability distribution of the reference point attribute according to the attribute context feature and the quantized reference point attribute includes: Processing the attribute context feature according to a second multi-layer perceptron to obtain Gaussian distribution parameters of each sub-attribute in the reference point attribute; The probability distribution of the reference point attribute is calculated according to the Gaussian distribution parameter, the quantized reference point attribute and the target quantization step size of the reference point attribute.

12. A 3D digital human and scene processing device, characterized in that: include: An acquisition module, used for acquiring 3D data based on three-dimensional Gaussian sputtering 3DGS; A representation generation module is used to generate a representation of the 3D data, specifically comprising: processing the 3D data to obtain a reference point position and a reference point attribute of each reference point, and performing feature transformation based on the reference point position and the reference point attribute in combination with the Gaussian sphere position generated corresponding to the reference point to obtain an attribute context feature of each reference point; A quantification module, configured to quantify the reference point attributes of the reference point using the attribute context feature; The entropy coding module is used to perform entropy coding on the quantized reference point attributes according to the attribute context features to obtain a coding result.

13. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, the steps of the method for processing 3D digital humans and scenes as described in any one of claims 1 to 11 are performed.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for processing a 3D digital human and a scene as claimed in any one of claims 1 to 11 are executed.

15. A computer program product, characterized in that The computer program product is stored in a storage medium, and the program product is executed by at least one processor to implement the steps of the method for processing 3D digital humans and scenes as claimed in any one of claims 1 to 11.

Images