A subjective quality assessment method for windowed six-degree-of-freedom composite videos

Abstract

The present invention specifically relates to a subjective quality evaluation method for a six-degree-of-freedom windowed synthetic video, including: building a corresponding subjective test environment based on the subjective quality evaluation requirements for the six-degree-of-freedom windowed synthetic video; selecting a number of source video sequences from a six-degree-of-freedom windowed multi-viewpoint plus depth video sequence to generate synthetic videos with different compression rates, video browsing paths, and reference viewpoint numbers; recruiting and training testers, obtaining the testers' subjective scores for each synthetic video in a subjective test environment, and calculating the corresponding average opinion scores to construct a subjective database; performing statistical analysis based on the average opinion scores to verify the rationality and effectiveness of the subjective database; and then evaluating the performance of the objective quality evaluation method through the subjective database to verify the necessity of establishing the subjective database. The present invention can effectively perceive the drawing distortion of the six-degree-of-freedom windowed synthetic video, and can adapt to the changes in the browsing path of the six-degree-of-freedom windowed synthetic video.

Description

Technical Field

[0001] The present invention relates to the technical field of video quality evaluation, and in particular to a subjective quality evaluation method for a window six-degree-of-freedom synthetic video. Background Art

[0002] Free viewpoint video allows users to switch viewpoints freely in the horizontal direction, improving the scene experience. With the rapid development of software and hardware technologies and users' pursuit of higher viewing experience, immersive video came into being. The Motion Picture Experts Group has formulated immersive video standards and is committed to promoting the development of sub-immersive video from three degrees of freedom to six degrees of freedom. Among them, window six-degree-of-freedom video allows users to watch at any angle on a plane, which is the future direction of immersive video technology development. In order to provide users with a comfortable and smooth viewpoint seamless switching viewing experience, the immersive video system needs to use drawing technology to generate virtual viewpoint synthetic video, but the drawing process will bring non-uniformly distributed local geometric distortion, that is, drawing distortion, which seriously affects the user's subjective perception.

[0003] Subjective quality evaluation is the most direct and accurate evaluation method that reflects human visual perception. Video subjective quality evaluation has received extensive attention from scholars in recent years. Xu Mai et al. proposed a panoramic video subjective quality evaluation method in the patent with publication number CN108055533B, which can effectively reflect the subjective quality of panoramic videos. However, the window six-degree-of-freedom synthetic video is different from the panoramic video. On the one hand, the window six-degree-of-freedom source video sequence is collected by a camera array arranged in a plane, and the field of view is extended to a two-dimensional space. It can move horizontally and vertically in the plane, and the introduced local drawing distortion is also derived to the vertical line. Its synthetic video can refer to the four adjacent source viewpoints, and the drawing distortion will increase with the decrease in the number of source reference viewpoints. Therefore, it is difficult for the existing subjective quality evaluation method to effectively perceive the drawing distortion of the window six-degree-of-freedom synthetic video. On the other hand, in an immersive experience, the human eye usually switches the viewing area autonomously following the dynamic changes of the video content. Different browsing paths may cause different judgments of the video quality by the observer, but the existing subjective quality evaluation methods are generally designed based on fixed browsing paths, which makes it difficult to evaluate the impact of browsing path changes on the subjective quality of the video. In summary, there is currently no subjective quality evaluation method specifically for windowed 6-DOF synthetic videos. Therefore, how to design a subjective quality evaluation method that can effectively perceive the drawing distortion of windowed 6-DOF synthetic videos and can adapt to changes in browsing paths is a technical problem that needs to be solved urgently. Summary of the invention

[0004] In view of the deficiencies of the above-mentioned prior art, the technical problem to be solved by the present invention is: how to provide a subjective quality evaluation method for a windowed six-degree-of-freedom synthetic video, so as to effectively perceive the drawing distortion of the windowed six-degree-of-freedom synthetic video and adapt to the changes in the browsing path of the windowed six-degree-of-freedom synthetic video, thereby improving the subjective quality evaluation effect of the windowed six-degree-of-freedom synthetic video and providing a reference for the research on the objective evaluation method of the windowed six-degree-of-freedom synthetic video.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0006] A subjective quality evaluation method for a window six-degree-of-freedom composite video comprises the following steps:

[0007] S1: Based on the subjective quality evaluation requirements of the window six-degree-of-freedom synthetic video, build the corresponding subjective test environment;

[0008] S2: Select several source video sequences from the windowed six-degree-of-freedom multi-view plus depth video sequence to generate synthetic videos with different compression rates, video browsing paths and reference viewpoint numbers;

[0009] S3: Recruit and train testers, obtain the testers' subjective scores on each synthetic video in a subjective test environment, calculate the corresponding average opinion scores, and build a subjective database;

[0010] S4: Perform statistical analysis based on the mean opinion score to verify the rationality and effectiveness of the subjective database; then evaluate the performance of the objective quality evaluation method through the subjective database to verify the necessity of establishing the subjective database.

[0011] Preferably, in step S2, the composite video is generated by the following steps:

[0012] S201: Select four source video sequences from a windowed six-degree-of-freedom multi-view plus depth video sequence;

[0013] S202: compressing the texture / depth videos of the four source video sequences in different combinations respectively;

[0014] S203: Generate videos of two browsing paths, namely, a fixed path and a significant path;

[0015] S204: Select one to four nearest reference viewpoints to draw synthetic viewpoints respectively, and generate corresponding synthetic videos.

[0016] Preferably, in step S201, the selected four source video sequences include two camera-shot videos and two computer-generated videos.

[0017] Preferably, in step S202, based on the characteristics of the texture video and the depth video, three QP combinations are compressed respectively: (25, 34), (35, 42) and (45, 48), where the former in brackets is the QP value of the texture video, and the latter is the QP value of the corresponding depth video; the compressed video together with the uncompressed source video sequence forms four corresponding texture / depth video combinations.

[0018] Preferably, in step S203, the fixed path is an S-shaped video browsing path formed by scanning from top to bottom and from left to right.

[0019] Preferably, in step S203, the saliency path is a saliency-related dynamic update path formed based on the dynamic capture of the foreground and motion changes of the video content by the human eye.

[0020] Preferably, when generating a saliency path, each frame of the video image is first segmented into superpixels, the edge probability is calculated, and the corresponding spatiotemporal edge probability map is obtained; then the foreground and background are segmented based on the geodesic distance, the saliency of each superpixel is calculated, and the centroid is determined according to the maximum value in the neighborhood; finally, the salient area of ​​the processed video sequence is updated every 20 frames, and the corresponding saliency path is drawn according to the change of the centroid of the salient area.

[0021] Preferably, according to the geodesic distance operator d g (.) Calculate the nth superpixel of the kth frame Significance

[0022]

[0023] Where: B represents background; B k , B k+1 Represent the background of the kth frame and the k+1th frame respectively; Represents an undirected weighted graph between adjacent frames;

[0024] Let S0 be the frame center, is the centroid of the salient area of ​​the reference view of the kth frame, is the focus point of the virtual view of the mth frame;

[0025] When drawing a saliency path, the path update is controlled within 20l, where l is the step length between adjacent frames; When the distance from S0 is not greater than 20l, Keep S0 unchanged; otherwise, focus on by The starting point is moved 20l in the direction of S0, that is, Where m=k+20.

[0026] Preferably, in step S3, the subjective scores of the testers are screened, the subjective scores of two unqualified testers are eliminated, and the average opinion score of each composite video is calculated:

[0027] in,

[0028] Where: MOS j represents the mean opinion score of the jth synthetic video; MOS ji represents the subjective score of the i-th tester on the j-th synthetic video; N represents the total number of testers.

[0029] Preferably, in step S4, the rationality, validity and necessity of the subjective database are verified by the following steps:

[0030] S401: Verify the validity of the subjective database based on the average opinion score distribution of each synthetic video;

[0031] S402: Perform principal component analysis and three-factor variance analysis based on the subjective scores to verify the rationality of the subjective database;

[0032] S403: Evaluate the performance of the objective quality evaluation method of the image / video through the subjective database to verify the necessity of establishing a subjective database for the window six-degree-of-freedom synthetic video.

[0033] The subjective quality evaluation method of the window six-degree-of-freedom composite video of the present invention has the following beneficial effects:

[0034] The present invention generates synthetic videos with different compression rates and reference viewpoints through multiple source video sequences, which can simulate the drawing distortion of the window six-degree-of-freedom synthetic video under immersive experience, so that the drawing distortion of the window six-degree-of-freedom synthetic video can be effectively perceived during subjective quality evaluation, and can adapt to the characteristics of the window six-degree-of-freedom synthetic video; at the same time, the present invention considers the different immersive experiences brought by different browsing behaviors of users by generating synthetic videos with different browsing paths, so that it can adapt to the browsing path changes of the window six-degree-of-freedom synthetic video, thereby improving the subjective quality evaluation effect of the window six-degree-of-freedom synthetic video. In addition, the present invention performs statistical analysis based on the average opinion score of the tester's subjective rating of the synthetic video, so that the user's immersive viewing experience can be comprehensively evaluated, and then the rationality and effectiveness of the subjective database can be verified; at the same time, by evaluating the performance of the existing advanced objective quality evaluation method on the subjective database, the necessity of establishing a subjective database for the window six-degree-of-freedom synthetic video can be verified, thereby better providing a reference for the research on the objective evaluation method of the window six-degree-of-freedom synthetic video. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to make the purpose, technical solution and advantages of the invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings, in which:

[0036] Figure 1 The logical block diagram of the subjective quality evaluation method for window six-degree-of-freedom synthetic video;

[0037] Figure 2 (a) is a schematic diagram of a fixed path. Figure 2 (b) is a schematic diagram of significant pathways;

[0038] Figure 3 (a) Figure 3 (b) Figure 3 (c) Figure 3 (d) The principal component analysis diagrams corresponding to MOS, QP, NV and NT respectively. DETAILED DESCRIPTION

[0039] The following is a further detailed description through specific implementation methods:

[0040] Example:

[0041] This embodiment discloses a subjective quality evaluation method for a window six-degree-of-freedom composite video.

[0042] like Figure 1 As shown, a subjective quality evaluation method for a window six-degree-of-freedom synthetic video includes the following steps:

[0043] S1: Based on the subjective quality evaluation requirements of the window six-degree-of-freedom synthetic video, build the corresponding subjective test environment;

[0044] S2: Select several source video sequences from the windowed six-degree-of-freedom multi-view video plus depth (MVD) video sequence to generate synthetic videos with different compression rates, video browsing paths and reference viewpoint numbers;

[0045] S3: Recruit and train testers, obtain their subjective scores on each synthetic video in a subjective test environment, calculate the corresponding mean opinion score (MOS), and build a subjective database;

[0046] S4: Conduct statistical analysis based on MOS values ​​to verify the rationality and effectiveness of the subjective database; then evaluate the performance of the objective quality evaluation method through the subjective database to verify the necessity of establishing the subjective database.

[0047] The present invention generates synthetic videos with different compression rates and reference viewpoints through multiple source video sequences, which can simulate the drawing distortion of the window six-degree-of-freedom synthetic video under immersive experience, so that the drawing distortion of the window six-degree-of-freedom synthetic video can be effectively perceived during subjective quality evaluation, and can adapt to the characteristics of the window six-degree-of-freedom synthetic video; at the same time, the present invention considers the different immersive experiences brought by different browsing behaviors of users by generating synthetic videos with different browsing paths, so that it can adapt to the browsing path changes of the window six-degree-of-freedom synthetic video, thereby improving the subjective quality evaluation effect of the window six-degree-of-freedom synthetic video. In addition, the present invention performs statistical analysis on the MOS value of the subjective score of the synthetic video by the tester, so that the user's immersive viewing experience can be comprehensively evaluated, and then the rationality and effectiveness of the subjective database can be verified; at the same time, by evaluating the performance of the existing advanced objective quality evaluation method on the subjective database, the necessity of establishing a subjective database for the window six-degree-of-freedom synthetic video can be verified, thereby better providing a reference for the research on the objective evaluation method of the window six-degree-of-freedom synthetic video.

[0048] In the specific implementation process, the subjective test environment built includes the test site and the software and hardware environment. Among them, the software and hardware environment are not limited to personal computer platforms or other higher-configuration mobile device platforms. The setting of the test site follows the relevant criteria for subjective video quality evaluation. It is necessary to select an indoor place with normal lighting and not easily disturbed, use an ultra-high-definition display device with a resolution not lower than that of the synthetic video, and the playback software can play smoothly without distortion.

[0049] In the specific implementation process, the synthetic video is generated through the following steps:

[0050] S201: Select four source video sequences from the windowed six-degree-of-freedom MVD video sequence; in this embodiment, two camera-shot videos (Technicolor Painter (TP) and ETRI Chef (EC)) and two computer-generated videos (Orange Shanman (OS) and Orange Kitchen (OK)) are selected.

[0051] S202: Using High Efficiency Video Coding (HEVC), the texture / depth videos of the four source video sequences are compressed in different combinations. In this embodiment, based on the characteristics of the texture video and the depth video, three QP combinations are compressed: (25, 34), (35, 42) and (45, 48), where the former in brackets is the QP value of the texture video and the latter is the QP value of the corresponding depth video. The compressed videos and the uncompressed source video sequences form four corresponding texture / depth video combinations, corresponding to four synthetic video quality levels of distortion-free (Lo), good (L1), medium (L2) and poor (L3).

[0052] S203: Generate videos of two navigation trajectories (NT), namely, a fixed path and a significant path. In this embodiment, Figure 2 The solid line in (a) is a fixed path (NT Fix ); Figure 2 The dashed line in (b) is the significant path (NT Sal ), the solid line is the saliency path actually drawn due to the limited field of view.

[0053] The fixed path is an S-shaped video browsing path formed by scanning from top to bottom and from left to right.

[0054] The saliency path is a dynamic update path related to saliency formed after the human eye dynamically captures the foreground and motion changes of the video content. When generating the saliency path, firstly, each frame of the video image is segmented into superpixels, the edge probability is calculated, and the corresponding spatiotemporal edge probability map is obtained; then the foreground and background are segmented based on the geodesic distance, the saliency of each superpixel is calculated, and the centroid is determined according to the maximum value in the neighborhood; finally, the salient area of ​​the processed video sequence is updated every 20 frames, and then the corresponding saliency path is drawn according to the change of the centroid of the salient area.

[0055] To ensure the comfort of human eyes, the step length between adjacent frames of the video sequence processed above is l, and in this embodiment, it is 4 to 6 pixels. g (.) Calculate the nth superpixel of the kth frame Significance

[0056]

[0057] In the formula: In the formula: B represents the background; B k , B k+1 Represent the background of the kth frame and the k+1th frame respectively; Represents an undirected weighted graph between adjacent frames;

[0058] Let S0 be the frame center, is the centroid of the salient area of ​​the reference view of the kth frame, is the focus point of the virtual view of the mth frame;

[0059] When drawing a saliency path, the path update is controlled within 20l, where l is the step length between adjacent frames; When the distance from S0 is not greater than 20l, Keep S0 unchanged; otherwise, focus on by The starting point is moved 20l in the direction of S0, that is, Where m=k+20.

[0060] in, It can be calculated as:

[0061]

[0062] in, for The distance between S0 and for to S0, l is the step size between adjacent frames.

[0063] S204: Select different numbers of reference viewpoints (Number of Viewpoint, NV) to draw synthetic viewpoints respectively, and generate corresponding synthetic videos. Based on the principle of proximity, this embodiment selects the nearest one to four reference viewpoints to draw synthetic viewpoints through the software VSRS4.3, and generates a total of 128 synthetic videos.

[0064] The present invention generates a synthetic video by using a fixed path and a saliency path, taking into account the different immersive experiences brought about by different browsing behaviors of users, so as to adapt to the browsing path changes of the window six-degree-of-freedom synthetic video, thereby improving the subjective quality evaluation effect of the window six-degree-of-freedom synthetic video. At the same time, the saliency path generated by the present invention can effectively simulate the attention area changes of the window six-degree-of-freedom synthetic video under the immersive experience, thereby further improving the subjective quality evaluation effect of the window six-degree-of-freedom synthetic video.

[0065] In the specific implementation process of step S3, it is necessary to recruit testers to conduct subjective tests and record subjective scores.

[0066] The testers must have qualified vision (or corrected vision), and it is recommended that there are no less than 20 testers, and the actual number of testers is 34. When scoring, the distorted video is played first for 10 seconds, and then a 5-second grayscale video is played for the testers to score. The five levels of perceived quality are Excellent, Good, Fair, Poor and Bad, corresponding to subjective scores of 5, 4, 3, 2 and 1 respectively.

[0067] Before the formal scoring, the testers were first informed of the experimental content and operation process, and trained through 5 "simulated demonstration videos". After passing the training, they took a short break (5 to 10 minutes), and then formally scored all the randomly played synthetic videos to avoid the training process affecting the formal test results.

[0068] Record each official scoring result for all test takers.

[0069] In the specific implementation process, the subjective scores of the testers were screened, the subjective scores of two unqualified testers were eliminated, and the MOS values ​​of each synthetic video were calculated:

[0070] in,

[0071] Where: MOS j represents the MOS value of the jth composite video; MOS ji represents the subjective score of the i-th tester on the j-th synthetic video; N represents the total number of testers, N=32.

[0072] During the specific implementation process, the rationality, effectiveness and necessity of the subjective database are verified through the following steps:

[0073] S401: Verify the validity of the subjective database based on the MOS value distribution of each synthetic video;

[0074] In this embodiment, the MOS value distribution of the synthetic sequence is statistically analyzed. According to the MOS value distribution diagram, it can be seen that the subjective scores of the testers for the synthetic videos generated by the significant path are generally higher than those for the synthetic videos generated by the fixed path, which verifies the effectiveness of the path design for the subjective quality evaluation of the window six-degree-of-freedom synthetic video.

[0075] S402: Perform principal component analysis (PCA) and three-way analysis of variance (ANOVA) based on the subjective scores to verify the rationality of the subjective database.

[0076] Specific: The principal component analysis diagram is as follows Figure 3 As shown; among them, Figure 3 (a) Figure 3 (b) Figure 3 (c) Figure 3(d) The principal component analysis diagrams corresponding to MOS, QP, NV and NT respectively. The principal component analysis diagram shows that PCA presents different shapes according to MOS, QP, NV and NT, showing a certain degree of differentiation. It can be seen that the MOS value, compression rate and number of reference viewpoints are clearly distinguished on the X-axis, indicating that they have a significant impact on the subjective results; the sequences of different paths can also be well distinguished on the Y-axis, and the MOS value of the sequence of the significant path is generally higher than that of the fixed path, that is, the video browsing path has an impact on subjective perception.

[0077] The ANOVA results are shown in Table 1, where Factors indicates the source of factors causing changes in MOS values, MS indicates mean square, DF indicates degrees of freedom, F value indicates the ratio of mean square of each effect term to error term, and P value indicates the statistical significance of the F test result. Usually, when P≤0.05, it indicates that the corresponding variable factors have significant differences. As can be seen from Table 1, the three-way ANOVA analysis takes into account the factors causing changes in MOS values. From the statistical significance of the F test result, P, it can be seen that the three basic variable factors NT, NV, and QP all have a significant impact on the MOS value (P≤0.05), and NT*NV or NT*QP do not have a significant interaction on the MOS value, indicating that the impact of the browsing path on the MOS value is independent of the compression rate or the number of reference viewpoints, which verifies the rationality of the subjective quality evaluation of the window six-degree-of-freedom synthetic video.

[0078] Table 1 Three-way ANOVA results

[0079]

[0080] S403: Evaluate the performance of the image / video objective quality evaluation method through the subjective database to verify the necessity of establishing the subjective database of the window six-degree-of-freedom synthetic video. In this embodiment, 9 advanced objective quality evaluation methods are used in the subjective database, and their objective scores in the original database and the current database are compared to verify the necessity of establishing the subjective database.

[0081] Nine non-reference objective indicators (BRISQUE, GM_LOG, VIIDEO, V-BLIINDS, NIQSV, NIQSV+, APT, MNSS and WANG's) were selected, including indicators for ordinary natural images or videos and indicators for DIBR synthetic images or videos. For image objective indicators, it is necessary to calculate frame by frame and then average all scores to obtain the objective score of the video. According to the recommendations of VQEGPhaseⅠFR-TV, the objective score and MOS value are fitted by nonlinear regression function to obtain the predicted subjective score MOS p :

[0082] MOS p =a·Q 3 +b·Q2 +c·Q+d;

[0083] Where: Q is the objective score, and parameters a, b, c, d are calculated by regression to minimize the difference between Q and MOS value.

[0084] The quality of an objective indicator is evaluated by calculating the correlation between the objective score and the MOS value. PLCC, SROCC and RMSE are often used to measure the accuracy, monotonicity and consistency of the objective indicator.

[0085] This embodiment tests 9 advanced image / video objective quality assessment methods in a subjective database, and compares the obtained results with the original results of each method, as shown in Table 2.

[0086] In Table 2, 'DB Ori 'The column contains the results from the original database, 'DB W6D The columns are the test results of the subjective database (W6D-VD) of the present invention.

[0087] Table 2 Performance comparison of 9 objective indicators in the original database and this database

[0088]

[0089] From Table 2, we can see that although the nine objective evaluation methods are Ori ' have achieved good performance, but in this database 'DB W6D ' performed poorly, with their PLCC and SROCC mostly below 0.6, indicating that they were not consistent with the subjective scores. The possible reasons are as follows: some indicators (GM_LOG and VIIDEO) focus on traditional compression distortion and perform poorly on DIBR synthetic videos; some indicators (GM_LOG, NIQSV, NIQSV+ and APT) are IQA indicators and cannot evaluate structure-related temporal distortion well; some indicators (NIQSV, NIQSV+ and APT) have achieved good results on the IRCCyN / IVC DIBR database and are relatively effective for some old-fashioned rendering distortions (such as large-area holes, stretching, etc.), but the above-mentioned old-fashioned distortions have mostly been well handled in existing algorithms. When such distortions are excluded from this database, these methods perform poorly; there are also a small number of indicators (NIQSV and APT) that directly ignore the impact of global compression on video perception; moreover, in new application scenarios, as NT changes, the distortion of structural information will affect the trajectory of the object viewed by the user and affect the perceived quality, and the above indicators do not take this factor into account. Therefore, it is necessary to establish a subjective database of the window six-degree-of-freedom synthetic video in the present invention.

[0090] The present invention can comprehensively evaluate the user's immersive viewing experience by statistically analyzing the testers' subjective scores and comparing them with the existing results of objective image / video quality evaluation, and further verify the rationality, effectiveness and necessity of the subjective database, thereby providing a better reference for the research on the objective evaluation method of window six-degree-of-freedom synthetic video.

[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit the technical solution. Those skilled in the art should understand that those modifications or equivalent substitutions of the technical solution of the present invention that do not depart from the purpose and scope of the technical solution should be included in the scope of the claims of the present invention.

Claims

1. A subjective quality evaluation method for a window six-degree-of-freedom synthetic video, characterized in that: The following steps are involved: S1: Based on the subjective quality evaluation requirements of the window six-degree-of-freedom synthetic video, build the corresponding subjective test environment; S2: Select several source video sequences from the windowed six-degree-of-freedom multi-view plus depth video sequence to generate synthetic videos with different compression rates, video browsing paths and reference viewpoint numbers; In step S2, a composite video is generated by the following steps: S201: Select four source video sequences from a windowed six-degree-of-freedom multi-view plus depth video sequence; S202: compressing the texture / depth videos of the four source video sequences in different combinations respectively; S203: Generate videos of two browsing paths, namely, a fixed path and a significant path; In step S203, the fixed path is an S-shaped video browsing path formed by scanning from top to bottom and from left to right; The saliency path is a dynamic update path related to saliency formed after the human eye dynamically captures the foreground and motion changes of the video content; When generating a saliency path, firstly, each frame of the video image is segmented into superpixels, the edge probability is calculated, and the corresponding spatiotemporal edge probability map is obtained; then, the foreground and background are segmented based on the geodesic distance, the saliency of each superpixel is calculated, and the centroid is determined according to the maximum value in the neighborhood; finally, the salient area of ​​the processed video sequence is updated every 20 frames, and then the corresponding saliency path is drawn according to the change of the centroid of the salient area; According to the geodesic distance operator d g (.) Calculate the nth superpixel of the kth frame Significance Where: B represents background; B k , B k+1 Represent the background of the kth frame and the k+1th frame respectively; Represents an undirected weighted graph between adjacent frames; Let S0 be the frame center, is the centroid of the salient area of ​​the reference view of the kth frame, is the focus point of the virtual view of the mth frame; When drawing a saliency path, the path update is controlled within 20l, where l is the step length between adjacent frames; When the distance from S0 is not greater than 20l, Keep S0 unchanged; otherwise, focus on by The starting point is moved 20l in the direction of S0, that is, Where m = k + 20; S204: Select one to four nearest reference viewpoints to draw synthetic viewpoints respectively, and generate corresponding synthetic videos; S3: Recruit and train testers, obtain the testers' subjective scores on each synthetic video in a subjective test environment, calculate the corresponding average opinion scores, and build a subjective database; S4: Perform statistical analysis based on the mean opinion score to verify the rationality and validity of the subjective database; Then, the performance of the objective quality evaluation method is evaluated through the subjective database to verify the necessity of establishing the subjective database.

2. The subjective quality evaluation method of a window six-degree-of-freedom composite video according to claim 1, characterized in that: In step S201, the four source video sequences selected include two camera-shot videos and two computer-generated videos.

3. The subjective quality evaluation method of a window six-degree-of-freedom composite video according to claim 1, characterized in that: In step S202, based on the characteristics of the texture video and the depth video, three QP combinations are compressed respectively: (25, 34), (35, 42) and (45, 48), where the former in brackets is the QP value of the texture video and the latter is the QP value of the corresponding depth video; the compressed video together with the uncompressed source video sequence forms four corresponding texture / depth video combinations.

4. The subjective quality evaluation method of a window six-degree-of-freedom composite video according to claim 1, characterized in that: Step S3, screen the subjective scores of the testers, remove the subjective scores of two unqualified testers, and calculate the average opinion score of each synthetic video: in, Where: MOS j represents the mean opinion score of the jth synthetic video; MOS ji represents the subjective score of the i-th tester on the j-th synthetic video; N represents the total number of testers.

5. The subjective quality evaluation method of a window six-degree-of-freedom composite video according to claim 1, characterized in that: In step S4, the rationality, validity and necessity of the subjective database are verified through the following steps: S401: Verify the validity of the subjective database based on the average opinion score distribution of each synthetic video; S402: Perform principal component analysis and three-factor variance analysis based on the subjective scores to verify the rationality of the subjective database; S403: Evaluate the performance of the objective quality evaluation method of the image / video through the subjective database to verify the necessity of establishing a subjective database for the window six-degree-of-freedom synthetic video.

Images