Method for encoding multi-view stitched video

Abstract

The application discloses a kind of multi-viewpoint splicing video encoding optimization method, comprising the following steps: S10, input the multi-viewpoint video sequence to be encoded, determine the number of viewpoints N to be spliced, and the key viewpoint K, the remaining viewpoints outside key viewpoint K are all auxiliary viewpoints A i , calculate the average offset of viewpoint Offset;S20, according to the number of viewpoints, viewpoint type, based on the shortest distance principle, all viewpoints of the previous frame in time domain are spliced into a large viewpoint frame, the number of key viewpoints is usually only one, the remaining N-1 viewpoints are all auxiliary viewpoints, and the distance d of the pixels in any auxiliary viewpoint and the pixels of the main viewpoint is the smallest shortest distance;S30, the video sequence composed of spliced large viewpoint frame is input into HEVC / H.265 standard encoder to perform encoding;S40, after all large viewpoint frames are encoded, output the code stream.

Description

Technical Field

[0001] This invention belongs to the field of video coding technology, specifically relating to a coding optimization method for multi-view spliced ​​video. Background Technology

[0002] In recent years, with the rapid development of multimedia technology, people are no longer satisfied with simple high-definition video applications, but hope to receive and experience video content from multiple perspectives through their own senses. This demand has promoted the development of ordinary single-view video content applications to multi-view video applications. Multi-view video refers to video recorded simultaneously by multiple cameras from different angles. In this type of video application, multi-view video compression is an important research and component. The international standards group has extended the single-view compression standard and released standards for multi-view video. Currently, the most advanced standard is the MV-HEVC standard, which is an extension of the HEVC standard. Compared with multicast coding, the technology used in the MV-HEVC standard can reduce the bitrate by 20% to 30% while maintaining the same image quality.

[0003] Unfortunately, in the current industry, standard codecs for single-view video still dominate the market. For cost and robustness reasons, people prefer to use these standard codecs to process various video content rather than purchasing separate codecs for multi-view video.

[0004] Therefore, establishing a method for processing multi-view video using standard codecs is crucial. The conventional approach in industry is to stitch together the viewpoints to be encoded (potentially side-by-side or top-to-bottom), and then treat the stitched video as a single stream for encoding and decoding. However, this method is overly complex and doesn't utilize the similarities between different viewpoints. To further reduce the bitrate of multi-view video compressed using this method, current research utilizes intra-block copying algorithms with screen content encoding to reduce information redundancy between different viewpoints, thereby minimizing the video bitrate. However, this method has two drawbacks: first, the intra-block copying algorithm has low accuracy and high complexity, making it unsuitable for industrial applications; second, it only provides a solution for two-viewpoint stitching but doesn't offer optimization methods for situations with more viewpoint stitching content. Summary of the Invention

[0005] In view of the above problems, the present invention provides an encoding optimization method for multi-view spliced ​​video, which uses a reasonable view splicing scheme and an optimized intra-frame block copying mode to significantly reduce the bitrate of encoding multi-view spliced ​​content using a standard encoder.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0007] A method for encoding optimization of multi-view stitched video includes the following steps:

[0008] S10, Input the multi-view video sequence to be encoded, determine the number of viewpoints N to be stitched together, and the key viewpoint K. All viewpoints other than the key viewpoint K are auxiliary viewpoints A. i Calculate the average viewpoint offset;

[0009] S20: Based on the number and type of viewpoints, and the principle of shortest distance, all viewpoints in the previous frame in the temporal domain are stitched together into a large viewpoint frame. There is usually only one key viewpoint, and the remaining N-1 viewpoints are all auxiliary viewpoints. The shortest distance means that the distance d between the pixel in any auxiliary viewpoint and the pixel in the key viewpoint is the smallest.

[0010] S30: The video sequence composed of stitched large viewpoint frames is input into the HEVC / H.265 standard encoder for encoding. Regardless of the number of stitched viewpoints, encoding begins from the starting position of the key viewpoint frame. The encoding order is based on raster scanning, and rate-distortion optimization is performed in units of CTU for each viewpoint. For the region where the key viewpoint frame is located, the intra-frame and inter-frame prediction during the encoding process is consistent with the standard encoding process. For the region where the auxiliary viewpoint frame is located, the intra-frame and inter-frame prediction involved in the standard encoding process are performed, as well as the intra-frame block copying mode based on the average viewpoint offset is executed to utilize the similarity between the key viewpoint and the auxiliary viewpoint for predictive encoding. For the boundary region where the key viewpoint and the auxiliary viewpoint are stitched together, additional sample adaptive compensation and deblocking filter techniques are performed after encoding to reduce edge discontinuities and block artifacts.

[0011] S40: After all large viewpoint frames are encoded, the output bitstream is completed.

[0012] In one possible implementation, the calculation of the average viewpoint offset in S10 specifically includes:

[0013] Perform a Discrete Fourier Transform (DFT) on all viewpoints in the first frame of the video sequence to convert the image from the spatial domain to the frequency domain. In the frequency domain, select the key viewpoint K and the auxiliary viewpoint A. i Calculate the Fourier transform results of these two images. and Calculate the ratio of their phase spectra:

[0014]

[0015] Where F represents the frequency domain representation of the signal, and u and v in F(u,v) represent the image signal in the horizontal and vertical directions, respectively. yes Complex conjugation, symbol Represents the magnitude of a complex number;

[0016] Then, a two-dimensional inverse Fourier transform (IDFT) is performed on this ratio to obtain the phase correlation function, also known as the dot product map. :

[0017]

[0018] The peak position of this dot product image represents the relative displacement between the two images. To find this position, we traverse the entire dot product image and find the position of its maximum value to obtain the relative offset between the two images:

[0019]

[0020] The relative offset of the i-th auxiliary viewpoint relative to the key viewpoint K is calculated. Then, the average relative offset Offset of the current frame is:

[0021]

[0022] This value reflects the average absolute offset between the key viewpoint and the secondary viewpoint, in pixels.

[0023] In one possible implementation, the intra-block copying mode based on the average viewpoint offset in S30 specifically includes:

[0024] Using the already encoded content within the current encoded frame as a reference region, inter-frame predictive coding is performed. Specifically, this can be described as follows: Assuming the encoded image frame is F, each block to be encoded can be labeled as... The target of intra-block copying is then represented as finding , so that:

[0025]

[0026] Since the pixel differences between image patches cannot be completely identical, the above equation is equivalent to:

[0027]

[0028] That is, to find the block that makes Dif as close to 0 as possible. If there are multiple such blocks, we further consider the distance between the two blocks and choose the block with the shortest distance as the final copy block. In other words, we want to find the block that requires the fewest bits to encode the block vector. This can be further expressed as the following optimization objective:

[0029]

[0030] Where BV represents the block vector bit(BV) represents the number of bits required to encode the block vector; These are the parameters used to determine the optimal block among multiple candidate blocks;

[0031] Considering that the encoded region will accumulate to a relatively large area at the end of the encoded frame, which increases the complexity of subsequent searches, the following fast search method is adopted: based on the location of the region processed by the current intra-block copy mode, find the location corresponding to the key viewpoint; based on this location information, add it to the Offset to obtain a widened region, which is the reference region for fast search optimization of intra-block copy;

[0032] After intra-frame block copying is completed, i.e., after obtaining the optimal block to be copied and its corresponding BV as described above, the predicted coding residual R and the searched block vector BV are encoded into the bitstream; the coding residual refers to the pixel difference between the block to be encoded and the copied block, i.e.:

[0033]

[0034] In terms of storage method, each auxiliary viewpoint only fully encodes the first BV, and the remaining BVs are stored in raster order, only encoding the difference between its BV and the first stored BV, in order to reduce the overall bit rate of the scheme as much as possible.

[0035] In one possible implementation, based on the location information, the area is added to the offset to obtain a widened region. Specifically, if the key viewpoint and the auxiliary viewpoint are spliced ​​left and right, then the horizontal component x is added to the offset.

[0036] In one possible implementation, based on the location information, the area is added to the offset to obtain a wider area. Specifically, if the key viewpoint and the auxiliary viewpoint are joined vertically, the vertical component y needs to be added to the offset.

[0037] In one possible implementation, based on the location information, the area is added to the offset to obtain a wider region. Specifically, if the key viewpoint and the auxiliary viewpoint are diagonally joined, then both the horizontal component x and the vertical component y are added to the offset.

[0038] The present invention has the following beneficial effects:

[0039] (1) The high-precision fast intra-block copy mode can make full use of the correlation between different viewpoints, thereby greatly reducing the complexity and bit rate of the coding auxiliary viewpoint, while reducing the coding complexity, making it more suitable for industrial applications.

[0040] (2) It can adapt to the splicing of N viewpoints and provides a variety of suitable splicing methods to minimize the complexity of encoding and save bit rate. Attached Figure Description

[0041] Figure 1 This is a flowchart illustrating the steps of an encoding optimization method for multi-view spliced ​​video according to an embodiment of the present invention.

[0042] Figure 2 This is a schematic diagram of the optimal stitching method with 2 viewpoints.

[0043] Figure 3 This is a schematic diagram of the optimal stitching method with 3 viewpoints.

[0044] Figure 4 This is a schematic diagram of the optimal stitching method with 4 viewpoints.

[0045] Figure 5 This is a schematic diagram of the optimal stitching method with 5 viewpoints.

[0046] Figure 6 This is a schematic diagram of the optimal stitching method with 7 viewpoints.

[0047] Figure 7 This is a schematic diagram illustrating the search area limitation for fast intra-frame block copying using a dual-viewpoint example.

[0048] Figure 8 This is a schematic diagram of a parallel encoding strategy. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] See Figure 1 The diagram shows a flowchart of a multi-viewpoint video stitching encoding optimization method according to an embodiment of the present invention, including the following steps:

[0051] S10, Input the multi-view video sequence to be encoded, based on the pre-set number of viewpoints N to be stitched together, and the key viewpoint K. All viewpoints other than the key viewpoint K are auxiliary viewpoints A. i The process calculates the average viewpoint offset. The key viewpoint serves as a reference point for auxiliary viewpoints; all analysis and predictions related to auxiliary viewpoints revolve around the key viewpoint. The number of viewpoints to be stitched together is predetermined. Viewpoint types include key viewpoints and auxiliary viewpoints. Key viewpoints, also called critical viewpoints, are also determined manually; all other viewpoints are auxiliary viewpoints.

[0052] S20: Based on the number and type of viewpoints, and the principle of shortest distance, all viewpoints in the previous frame in the temporal domain are stitched together into a large viewpoint frame. There is usually only one key viewpoint, and the remaining N-1 viewpoints are all auxiliary viewpoints. The shortest distance means that the distance d between the pixel in any auxiliary viewpoint and the pixel in the key viewpoint is the smallest.

[0053] S30: The video sequence composed of stitched large viewpoint frames is input into the HEVC / H.265 standard encoder for encoding. Regardless of the number of stitched viewpoints, encoding begins from the starting position of the key viewpoint frame. The encoding order is based on raster scanning, and rate-distortion optimization is performed in units of CTU for each viewpoint. For the region where the key viewpoint frame is located, the intra-frame and inter-frame prediction during the encoding process is consistent with the standard encoding process. For the region where the auxiliary viewpoint frame is located, the intra-frame and inter-frame prediction involved in the standard encoding process are performed, as well as the intra-frame block copying mode based on the average viewpoint offset is executed to utilize the similarity between the key viewpoint and the auxiliary viewpoint for predictive encoding. For the boundary region where the key viewpoint and the auxiliary viewpoint are stitched together, additional sample adaptive compensation and deblocking filter techniques are performed after encoding to reduce edge discontinuities and block artifacts.

[0054] Intra-frame prediction refers to prediction within a single frame, primarily used to reduce spatial redundancy within the same frame. Specifically, it uses adjacent pixel blocks (macroblocks or coded blocks) already encoded in the current frame to predict the content of uncoded blocks. By predicting similar regions, the need to store or transmit each pixel can be reduced. Inter-frame prediction, on the other hand, utilizes the temporal redundancy of the video, that is, it uses the similarity between consecutive frames for prediction. It reduces temporal redundancy by referencing other frames (usually previous or subsequent frames) to predict pixel blocks in the current frame. Inter-frame prediction can significantly improve compression efficiency because adjacent frames in a video are often very similar in content.

[0055] The Sample Adaptive Offset (SAO) process consists of two stages: classification and compensation. In the classification stage, SAO divides pixels into different categories and adjusts the offset values ​​for each category. The purpose of classification is to find pixels with similar characteristics (such as edges, flat areas) and calculate appropriate compensation for them. SAO provides two classification methods: edge classification and bandwidth classification. The former classifies pixels into different edge directions (such as horizontal, vertical, diagonal, etc.) by examining the directional changes in the pixel's surrounding neighborhood; the latter divides pixel values ​​according to predefined intervals. For example, pixel values ​​can be divided into multiple bandwidths based on brightness values, and pixels in each bandwidth can use different offset values. In the compensation stage, SAO calculates the offset value for each category or bandwidth based on the results of the classification stage and applies these offset values ​​to the pixels of the reconstructed image. The purpose of the offset values ​​is to correct the quantization error of the reconstructed image, making it closer to the original image. The deblocking filter (DF) process mainly consists of three stages. The first stage is to detect distortion at block boundaries and determine which boundaries need smoothing. The second stage applies adaptive filtering. After detecting the block boundaries that need to be processed, the filter selects different filtering intensities based on the characteristics of the block boundaries (such as whether they belong to high-frequency regions and the boundary strength). The filter strength typically depends on several factors: quantization parameters, block motion vectors, and pixel differences within and between blocks. Finally, DF adjusts the pixel values ​​near the block boundaries to make the transition between adjacent blocks smoother. It typically processes the vertical and horizontal boundaries of the blocks separately.

[0056] S40: After all large viewpoint frames are encoded, the output bitstream is completed.

[0057] A specific application example of the present invention provides a method for encoding optimization of multi-viewpoint stitched video. In step S10, calculating the average viewpoint offset specifically includes:

[0058] Perform a Discrete Fourier Transform (DFT) on all viewpoints in the first frame of the video sequence to convert the image from the spatial domain to the frequency domain. In the frequency domain, select the key viewpoint K and the auxiliary viewpoint A. i Calculate the Fourier transform results of these two images. and Calculate the ratio of their phase spectra:

[0059]

[0060] Where F represents the frequency domain representation of the signal, and u and v in F(u,v) represent the image signal in the horizontal and vertical directions, respectively. yes Complex conjugation, symbol Represents the magnitude of a complex number;

[0061] Then, a two-dimensional inverse Fourier transform (IDFT) is performed on this ratio to obtain the phase correlation function, also known as the dot product map. :

[0062]

[0063] The peak position of this dot product image represents the relative displacement between the two images. To find this position, we traverse the entire dot product image and find the position of its maximum value to obtain the relative offset between the two images:

[0064]

[0065] The relative offset of the i-th auxiliary viewpoint relative to the key viewpoint K is calculated. Then, the average relative offset Offset of the current frame is:

[0066]

[0067] This value reflects the average absolute offset between the key viewpoint and the auxiliary viewpoint, in pixels. Note that this embodiment of the invention applies only to scenes where the camera's shooting angle changes little or during translational motion; it is not applicable to other scenes with larger changes. Generally, "little change in shooting angle" means that the shooting angle between adjacent cameras is less than or equal to 15°, and "larger" means greater than 15°.

[0068] A specific application example of the present invention provides an encoding optimization method for multi-view spliced ​​video. In S20, Figures 2 to 6 The optimal stitching methods are shown for 2, 3, 4, 5, and 7 viewpoints. All are based on the aforementioned shortest distance principle, minimizing the distance between as many auxiliary viewpoints as possible and the key viewpoint. Viewpoint 0 is the key viewpoint, and the rest are auxiliary viewpoints. The dashed lines are used to complete the image and are filled with gray pixels; they do not participate in the actual encoding and decoding. The distance d is calculated as follows:

[0069]

[0070] Obviously, the pixel distance d between two blocks is minimized when the distance between any two pixels is the minimum of their width and height.

[0071] A specific application example of the present invention provides an encoding optimization method for multi-viewpoint stitched video. In step S30, the execution of an intra-block copying mode based on the average viewpoint offset specifically includes:

[0072] Using the already encoded content within the current encoded frame as a reference region, inter-frame predictive coding is performed. In this invention, the intra-frame block copy mode only applies to the rate-distortion optimization process of auxiliary viewpoints, and its reference region is the key viewpoint that has already been encoded. To reduce the complexity of this mode, the reference region is limited based on the Offset calculated in step 1, such as... Figure 7 As shown, taking a dual-viewpoint example, to limit the search area for fast intra-frame block copying, based on the region information at position (x, y), its horizontal component x is added to the offset to obtain a widened region (x + Offset, y). It's important to note that if the key viewpoint and auxiliary viewpoint are joined vertically, the vertical component y needs to be added to the offset; if the key viewpoint and auxiliary viewpoint are joined diagonally, both the horizontal component x and the vertical component y need to be added to the offset.

[0073] Assume the encoded image frame is F, where each block to be encoded can be labeled as The target of intra-block copying is then represented as finding , so that:

[0074]

[0075] Since the pixel differences between image patches cannot be completely identical, the above equation is equivalent to:

[0076]

[0077] That is, to find the block that makes Dif as close to 0 as possible. If there are multiple such blocks, we further consider the distance between the two blocks and choose the block with the shortest distance as the final copy block. In other words, we want to find the block that requires the fewest bits to encode the block vector. This can be further expressed as the following optimization objective:

[0078]

[0079] Where BV represents the block vector bit(BV) represents the number of bits required to encode the block vector; These are the parameters used to determine the optimal block among multiple candidate blocks;

[0080] Considering that the encoded region will accumulate to a relatively large area at the end of the encoded frame, which increases the complexity of subsequent searches, the following fast search method is adopted: based on the location of the region processed by the current intra-block copy mode, find the location corresponding to the key viewpoint; based on this location information, add it to the Offset to obtain a widened region, which is the reference region for fast search optimization of intra-block copy;

[0081] After intra-frame block copying is completed, i.e., after obtaining the optimal block to be copied and its corresponding BV as described above, the predicted coding residual R and the searched block vector BV are encoded into the bitstream; the coding residual refers to the pixel difference between the block to be encoded and the copied block, i.e.:

[0082]

[0083] In terms of storage method, each auxiliary viewpoint only fully encodes the first BV, and the remaining BVs are stored in raster order, only encoding the difference between its BV and the first stored BV, in order to reduce the overall bit rate of the scheme as much as possible.

[0084] Furthermore, in the application examples of this invention, a parallel processing strategy is also used to reduce the algorithm complexity. For example... Figure 8 As shown, once the key viewpoint has completed two lines of encoding, its auxiliary viewpoint can begin encoding.

[0085] Table 1 shows the bitrate savings and encoding speed changes of dual-view and tri-view test videos after applying the optimization algorithm.

[0086]

[0087] To verify the results of the method in this embodiment of the invention, considering that most multi-view video sources are currently dual-view or triple-view videos, Table 1 shows the bitrate savings and encoding speed changes of dual-view and triple-view test videos after applying the optimization algorithm. The comparison object is the encoding results of these video sources in the standard codec. The BD-Rate index represents the bitrate savings, with negative values ​​representing the percentage of bitrate savings under the same objective quality. It can be clearly observed that, regardless of whether in All-Intra or RandomAccess mode, the optimization scheme of the method in this embodiment of the invention has a significant bitrate improvement effect.

[0088] It should be understood that the exemplary embodiments described herein are illustrative and not restrictive. Although one or more embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art will understand that various changes in form and detail may be made without departing from the spirit and scope of the invention as defined by the appended claims.

Claims

1. A method for optimizing the encoding of multi-view stitched video, characterized in that, Includes the following steps: S10, Input the multi-view video sequence to be encoded, determine the number of viewpoints N to be stitched together, and the key viewpoint K. All viewpoints other than the key viewpoint K are auxiliary viewpoints A. i Calculate the average viewpoint offset, which reflects the average absolute offset between the key viewpoint and the auxiliary viewpoint, in pixels. S20: Based on the number and type of viewpoints, and the principle of shortest distance, all viewpoints in the previous frame in the temporal domain are stitched together into a large viewpoint frame. There is only one key viewpoint, and the remaining N-1 viewpoints are all auxiliary viewpoints. The shortest distance means that the distance d between the pixel in any auxiliary viewpoint and the pixel in the key viewpoint is the smallest. S30 inputs the video sequence composed of spliced ​​large viewpoint frames into the HEVC / H.265 standard encoder for encoding. Encoding starts from the beginning position of the key viewpoint frame. The encoding order is based on raster scanning. Rate distortion optimization is performed in CTU units in each viewpoint. For the area where the key viewpoint frame is located, the intra-frame and inter-frame prediction in the encoding process is consistent with the standard encoding process. For the region where the auxiliary viewpoint frame is located, intra-frame and inter-frame predictions are performed as part of the standard coding process, and an intra-block copy mode based on the average viewpoint offset is executed to utilize the similarity between the key viewpoint and the auxiliary viewpoint for predictive coding. For the boundary region where the key viewpoint and the auxiliary viewpoint are stitched together, additional sample point adaptive compensation and deblocking filter techniques are performed after encoding to reduce edge discontinuities and block effects. S40: After all large viewpoint frames are encoded, the output bitstream is completed.

2. The encoding optimization method for multi-view spliced ​​video as described in claim 1, characterized in that, The calculation of the average viewpoint offset described in S10 specifically includes: Perform a Discrete Fourier Transform (DFT) on all viewpoints in the first frame of the video sequence to convert the image from the spatial domain to the frequency domain; in the frequency domain, select the key viewpoint K and the auxiliary viewpoint A. i Calculate the Fourier transform results of these two images. and Calculate the ratio of their phase spectra: Where F represents the frequency domain representation of the signal, and u and v in F(u,v) represent the image signal in the horizontal and vertical directions, respectively. yes Complex conjugation, symbol Represents the magnitude of a complex number; Then, a two-dimensional inverse Fourier transform (IDFT) is performed on this ratio to obtain the phase correlation function, also known as the dot product map. : The peak position of this dot product image represents the relative displacement between the two images. To find this position, we traverse the entire dot product image and find the position of its maximum value to obtain the relative offset between the two images: The relative offset of the i-th auxiliary viewpoint to the key viewpoint K is calculated, and the average viewpoint offset Offset of the current frame is: 。 3. The encoding optimization method for multi-view spliced ​​video as described in claim 2, characterized in that, The intra-block copying mode based on the average viewpoint offset in S30 specifically includes: Using the already encoded content within the current encoded frame as a reference region, inter-frame predictive coding is performed. Specifically, this can be described as follows: the encoded image frame is F, where each block to be encoded can be labeled as... The target of intra-block copying is then represented as finding , so that: Since the pixel differences between image patches cannot be completely identical, the above equation is equivalent to: That is, to find the block that makes Dif as close to 0 as possible. If there are multiple such blocks, we further consider the distance between the two blocks and choose the block with the shortest distance as the final copy block, which is the block that requires the fewest bits to encode the block vector. This can be further expressed as the following optimization objective: Where BV represents the block vector bit(BV) represents the number of bits required to encode the block vector; These are the parameters used to determine the optimal block among multiple candidate blocks; Considering that the encoded region will accumulate to a relatively large area at the end of the encoded frame, which increases the complexity of subsequent searches, the following fast search method is adopted: based on the location of the region processed by the current intra-block copy mode, find the location corresponding to the key viewpoint; based on this location information, add it to the Offset to obtain a widened region, which is the reference region for fast search optimization of intra-block copy; After intra-frame block copying is completed, i.e., after obtaining the optimal block to be copied and its corresponding BV as described above, the predicted coding residual R and the searched block vector BV are encoded into the bitstream; the coding residual refers to the pixel difference between the block to be encoded and the copied block, i.e.: In terms of storage method, each auxiliary viewpoint only fully encodes the first BV, and the remaining BVs are stored in raster order, only encoding the difference between its BV and the first stored BV.

4. The encoding optimization method for multi-view spliced ​​video as described in claim 3, characterized in that, Based on this location information, it is added to the Offset to obtain an expanded area. Specifically, if the key viewpoint and the auxiliary viewpoint are spliced ​​left and right, then its horizontal component x is added to the Offset.

5. The encoding optimization method for multi-view spliced ​​video as described in claim 3, characterized in that, Based on this location information, it is added to the Offset to obtain an expanded area. Specifically, if the key viewpoint and the auxiliary viewpoint are joined vertically, the vertical component y needs to be added to the Offset.

6. The encoding optimization method for multi-view spliced ​​video as described in claim 3, characterized in that, Based on this location information, it is added to the Offset to obtain an expanded area. Specifically, if the key viewpoint and the auxiliary viewpoint are diagonally joined, then both the horizontal component x and the vertical component y are added to the Offset.

Images