A bit rate allocation method for dynamic point cloud coding based on perceptual analysis

Abstract

The present invention discloses a dynamic point cloud coding rate allocation method based on perceptual analysis, which obtains a total coding rate model of the dynamic point cloud according to the respective coding rate models of geometric information and color information; establishes respective quantized subjective quality models of geometric information and color information to obtain a quantized subjective quality model of the dynamic point cloud; determines a traversal interval of the color coding quantization parameter, calculates the predicted total coding rate of the dynamic point cloud under each group of coding quantization parameter groups according to the total coding rate model of the dynamic point cloud within the traversal interval, and adds the coding quantization parameter groups that meet the conditions into a candidate list; calculates the quantized subjective quality of the dynamic point cloud under each group of coding quantization parameter groups in the candidate list according to the quantized subjective quality model of the dynamic point cloud; adopts the coding quantization parameter group corresponding to the maximum quantized subjective quality as the optimal coding quantization parameter group for encoding; the advantage is that the subjective quality is guaranteed while accurately allocating the point cloud geometry and color coding rate.

Description

Technical Field

[0001] The present invention relates to a dynamic point cloud coding technology, and in particular to a dynamic point cloud coding bit rate allocation method based on perceptual analysis. Background Art

[0002] With the rapid development of computer science and technology, people have higher and higher requirements for the experience of multimedia applications, and multimedia applications have entered a new era of development. As an efficient form of immersive media, point cloud is used in many fields such as virtual reality, medical imaging, and autonomous driving. However, the huge amount of point cloud data seriously affects its wider application. In order to effectively compress dynamic point cloud data, the Moving Picture Experts Group (MPEG) proposed the Video-Based Point Cloud Compression (V-PCC) standard. The main idea of ​​the V-PCC standard is to project the geometry and color information of the dynamic point cloud into the projection video (geometry video and color video) and occupancy map sequence, and then use advanced video encoders to compress the projection video. At present, some related studies on the bit rate allocation of dynamic point cloud coding for the V-PCC standard have good bit error rate control, but do not fully consider the subjective quality of reconstructed dynamic point clouds. There are also some methods that maintain good geometric quality of reconstructed dynamic point clouds, but their color quality is seriously distorted. This is because the dynamic point cloud distortion measurement model they use is not suitable. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a dynamic point cloud coding rate allocation method based on perceptual analysis, which ensures subjective quality while accurately allocating point cloud geometry and color coding rates.

[0004] The technical solution adopted by the present invention to solve the above technical problems is: a dynamic point cloud coding rate allocation method based on perceptual analysis, characterized by comprising the following steps:

[0005] Step 1: Under the V-PCC standard, the encoding rate of the geometric information of the dynamic point cloud is recorded as R g , and described as: The encoding rate of the color information of the dynamic point cloud is recorded as R c , and described as: Among them, α g and β g is the geometric encoding parameter, Q g represents the geometric coding quantization step size, α c and β c is the color coding parameter, Q cIndicates the color coding quantization step size;

[0006] Step 2: Quantize step size Q in geometric coding g The geometric information of the dynamic point cloud is encoded when the values ​​are 3.25, 6.5, 16, and 40, respectively, corresponding to the actual encoding bit rate of the geometric information of the dynamic point cloud; then, according to the geometric coding quantization step size Q g The four values ​​of and the corresponding four actual encoding rates are Perform the least squares fitting and get α g and β g The value of

[0007] Similarly, the color coding quantization step size Q c The color information of the dynamic point cloud is encoded when the values ​​are 16, 32, and 64 respectively, corresponding to the actual encoding bit rate of the color information of the dynamic point cloud; then according to the color coding quantization step size Q c The three values ​​of and the corresponding three actual encoding rates are Perform the least squares fitting and get α c and β c The value of

[0008] Step 3: Based on step 2, and Get the encoding quantization step size group (Q g ,Q c ) represents the total encoding rate of the dynamic point cloud, denoted by R(Q g ,Q c ) and described as: Then The geometric coding quantization step size Q in g Convert to geometric coding quantization parameter QP g , color-coded quantization step size Q c Convert to color encoding quantization parameter QP c , get the encoding quantization parameter group (QP g ,QP c ) represents the total encoding bit rate of the dynamic point cloud, denoted as R(QP g ,QP c ) and described as: Wherein, f() represents the process of converting the coding quantization parameter to the coding quantization step size;

[0009] Step 4: Geometry-based coding quantization parameter QP g A quantitative subjective quality model of the geometric information of dynamic point clouds is established, which is described as: And based on the color encoding quantization parameter QP cA quantitative subjective quality model of the color information of dynamic point clouds is established, which is described as: Among them, S g (QP g ) represents the quantization parameter QP based on geometric coding g Quantifying the subjective quality of the geometric information of a dynamic point cloud, S m Indicates the highest quantized subjective quality set, QP g,v Indicates the set geometric coding quantization parameter threshold, S c (QP c ) represents the color-coded quantization parameter QP c The quantified subjective quality of the color information of the dynamic point cloud, QP c,v Indicates the set color-coded quantization parameter threshold;

[0010] Step 5: Based on step 4, and Get the encoding quantization parameter group (QP g ,QP c ) is represented by the quantized subjective quality model of the dynamic point cloud, which is described as: S(QP g ,QP c )=S g (QP g )+S c (QP c ); where S(QP g ,QP c ) indicates the encoding quantization parameter group (QP g ,QP c ) represents the quantified subjective quality of the dynamic point cloud;

[0011] Step 6: Determine the geometric coding quantization parameter QP g The termination value is denoted as QP g,final , And determine the color encoding quantization parameter QP c The starting value is denoted as QP c,start , Among them, f -1 () represents the process of converting the coding quantization step size to the coding quantization parameter, R target represents the target encoding bit rate to be allocated;

[0012] Step 7: Traverse the interval [QP c,start ,51], define the currently traversed value as the current value;

[0013] Step 8: Use the current value as the current predicted color encoding quantization parameter and record it as QP c,cur ; Then according to QP c,curEncode the color information of the dynamic point cloud to obtain the predicted coding rate of the color information of the dynamic point cloud, denoted as R c,cur , Then calculate the remaining bit rate, denoted as R res , R res =R target -R c,cur ; Then calculate the current predicted color coding quantization parameter QP c,cur The corresponding current prediction geometry coding quantization parameter is denoted as QP g,cur , Finally, according to step 3 Calculate the encoding quantization parameter set (QP g,cur ,QP c,cur ) represents the predicted total coding rate of the dynamic point cloud, denoted by R(QP g,cur ,QP c,cur ), Among them, QP c,cur ∈[QP c,start ,51],QP g,cur ∈[QP g,final ,51];

[0014] Step 9: Determine R(QP g,cur ,QP c,cur ) Whether the conditions are met If satisfied, the quantization parameter group (QP g,cur ,QP c,cur ) is added to the candidate list, and then step 10 is performed; otherwise, the encoding quantization parameter group (QP g,cur ,QP c,cur ), and then execute step 10; wherein the symbol “||” is an absolute value symbol, T represents the limit bit error rate, T∈[0,1];

[0015] Step 10: Set the interval [QP c,start ,51] takes the next traversal value as the current value, and then returns to step 8 to continue until the interval [QP c,start ,51], each value in the traversal is completed, and the candidate list is obtained;

[0016] Step 11: According to S(QP in step 5 g ,QP c )=S g (QP g )+S c (QP c ), calculate the quantized subjective quality of the dynamic point cloud represented by each set of coding quantization parameter groups in the candidate list, and take the kth set of coding quantization parameter groups (QP g,k ,QP c,kThe quantized subjective quality of the dynamic point cloud represented by ) is denoted as S(QP g,k ,QP c,k ), S(QP g,k ,QP c,k )=S g (QP g,k )+S c (QP c,k ), where 1≤k≤K, K represents the number of coding quantization parameter groups included in the candidate list, S g (QP g,k )pass Calculated, S c (QP c,k )pass Calculated;

[0017] Step 12: Find the quantized subjective quality with the largest value from the K quantized subjective qualities calculated in step 11; then use a set of coding quantization parameter groups corresponding to the quantized subjective quality with the largest value as a set of coding quantization parameter groups with the best subjective quality; and then encode the dynamic point cloud according to the set of coding quantization parameter groups with the best subjective quality.

[0018] In step 4, S m =51,QP g,v =10,QP c,v =28.

[0019] In the step 9, T=0.03.

[0020] Compared with the prior art, the advantages of the present invention are:

[0021] The method of the present invention firstly establishes a coding rate model of geometric information based on the geometric coding quantization step, and establishes a coding rate model of color information based on the color coding quantization step, and obtains model parameters by selecting reasonable geometric coding quantization step and color coding quantization step to fit, and then obtains the total coding rate model of the dynamic point cloud represented by the coding quantization parameter group; secondly, in order to solve the problem of unreasonable dynamic point cloud quality evaluation in the process of dynamic point cloud rate allocation, and to make the parameters of the quantitative subjective quality model and the coding rate model consistent, the influence of different geometric coding quantization parameters and color coding quantization parameters on the subjective quality of the reconstructed point cloud is analyzed, and the quantitative subjective quality model of the dynamic point cloud is established based on the geometric coding quantization parameters and the color coding quantization parameters; finally, in order to quickly find a set of optimal coding quantization parameter groups, a method for determining the optimal coding quantization parameter group under the extreme bit error rate is proposed, and the method can also more accurately allocate the bit rate for the geometric information and color information of the dynamic point cloud. Experimental results show that the subjective quality of the reconstructed point cloud using the method of the present invention is good and the bit error rate is low. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 The figure is a general implementation block diagram of the method of the present invention. DETAILED DESCRIPTION

[0023] The present invention is further described in detail below with reference to the accompanying drawings.

[0024] The present invention proposes a method for allocating bit rate of dynamic point cloud coding based on perceptual analysis, and its overall implementation block diagram is as follows: Figure 1 As shown, it includes the following steps:

[0025] Step 1: Under the V-PCC standard, the encoding rate of the geometric information of the dynamic point cloud is recorded as R g , and described as: The encoding rate of the color information of the dynamic point cloud is recorded as R c , and described as: Among them, α g and β g is the geometric encoding parameter, Q g represents the geometric coding quantization step size, α c and β c is the color coding parameter, Q c Indicates the color encoding quantization step size.

[0026] Step 2: Quantize step size Q in geometric coding g The geometric information of the dynamic point cloud is encoded when the values ​​are 3.25, 6.5, 16, and 40, respectively, corresponding to the actual encoding bit rate of the geometric information of the dynamic point cloud; then, according to the geometric coding quantization step size Q g The four values ​​of and the corresponding four actual encoding rates are Perform the least squares fitting and get α g and β g The value of .

[0027] Similarly, the color coding quantization step size Q c The color information of the dynamic point cloud is encoded when the values ​​are 16, 32, and 64 respectively, corresponding to the actual encoding bit rate of the color information of the dynamic point cloud; then according to the color coding quantization step size Q c The three values ​​of and the corresponding three actual encoding rates are Perform the least squares fitting and get α c and β c The value of .

[0028] Step 3: Based on step 2, and Get the encoding quantization step size group (Q g ,Q c) represents the total encoding rate of the dynamic point cloud, denoted by R(Q g ,Q c ) and described as: Then The geometric coding quantization step size Q in g Convert to geometric coding quantization parameter QP g , color-coded quantization step size Q c Convert to color encoding quantization parameter QP c , get the encoding quantization parameter group (QP g ,QP c ) represents the total encoding bit rate of the dynamic point cloud, denoted as R(QP g ,QP c ) and described as: Among them, f() represents the process of converting the coding quantization parameter to the coding quantization step size.

[0029] Step 4: Geometry-based coding quantization parameter QP g A quantitative subjective quality model of the geometric information of dynamic point clouds is established, which is described as: And based on the color encoding quantization parameter QP c A quantitative subjective quality model of the color information of dynamic point clouds is established, which is described as: Among them, S g (QP g ) represents the quantization parameter QP based on geometric coding g Quantifying the subjective quality of the geometric information of a dynamic point cloud, S m Indicates the highest quantized subjective quality set, which is S in this embodiment. m =51,QP g,v Indicates the set geometric coding quantization parameter threshold. In this embodiment, QP g,v =10, S c (QP c ) represents the color-coded quantization parameter QP c The quantified subjective quality of the color information of the dynamic point cloud, QP c,v Indicates the color coding quantization parameter threshold value set. In this embodiment, QP c,v =28.

[0030] The existing dynamic point cloud distortion measurement model has the problem of unreasonable weighting. To solve this problem, a quantitative subjective quality model of the geometric information of the dynamic point cloud and a quantitative subjective quality model of the color information of the dynamic point cloud are established at the same level, and the QP is explained through subjective analysis. g and QP cImpact on the subjective quality of the reconstructed point cloud. 10 people (5 males and 5 females) were selected to participate in the subjective scoring experiment. They had no professional background and normal vision. In the subjective scoring experiment, the original point cloud was located on the left, the reference point cloud was located in the middle, and the comparison point cloud was located on the right. The distortion of the point cloud geometry and color was analyzed respectively. The original point cloud was used as the initial reference point cloud, and the reconstructed point cloud when the encoding quantization parameter QP was 1 was used as the initial comparison point cloud. When the distortion of the comparison point cloud was not perceived to be more serious than that of the reference point cloud, the comparison point cloud was replaced by the reconstructed point cloud when QP = QP + ΔQP, ΔQP = 2, and recorded as 0; otherwise, the reference point cloud was replaced by the current comparison point cloud and recorded as -1. The above process was repeated until the encoding quantization parameter of the comparison point cloud was 51. The quantized subjective quality analysis curve of geometry and color is divided into three sections, which remain horizontal in the initial section, decrease in a ripple shape in the middle section, and finally decrease linearly. The initial segment remains horizontal, indicating that when the coding quantization parameter is small, the coding distortion generated is not enough to cause changes in the quantized subjective quality; the middle segment shows a ripple-like decline, indicating that the coding distortion of consecutive coding quantization parameters is not much different and difficult for the human eye to distinguish; the last segment shows a linear decline, indicating that when the coding quantization parameter is large, the coding distortion of the previous and next coding quantization parameters is different and can be distinguished by the human eye. Through the above perceptual analysis, in order to simplify the relationship between the quantized subjective quality and the coding quantization parameter, the three-segment type is adjusted to a two-segment type (i.e. ) to represent the quantified subjective quality of dynamic point cloud geometry and color.

[0031] Step 5: Based on step 4, and Get the encoding quantization parameter group (QP g ,QP c ) is represented by the quantized subjective quality model of the dynamic point cloud, which is described as: S(QP g ,QP c )=S g (QP g )+S c (QP c ); where S(QP g ,QP c ) indicates the encoding quantization parameter group (QP g ,QP c ) to quantify the subjective quality of dynamic point clouds represented by .

[0032] Here, the model S(QP g ,QP c )=S g (QP g )+S c (QP c) quantifies the subjective quality of the geometric information and color information of the dynamic point cloud to the same level, effectively solving the problem of unreasonable weighting of geometric and color distortions, and its model parameters are consistent with the bit rate model and are compatible with the bit rate model.

[0033] Step 6: Determine the geometric coding quantization parameter QP g The termination value is denoted as QP g,final , And determine the color encoding quantization parameter QP c The starting value is denoted as QP c,start , Among them, f -1 () represents the process of converting the coding quantization step size to the coding quantization parameter, R target Indicates the target encoding bit rate to be allocated, where the geometric encoding quantization parameter QP is determined g The minimum value and color encoding quantization parameter QP c The minimum value of .

[0034] Step 7: Traverse the interval [QP c,start ,51], defines the currently traversed value as the current value.

[0035] Step 8: Use the current value as the current predicted color encoding quantization parameter and record it as QP c,cur ; Then according to QP c,cur Encode the color information of the dynamic point cloud to obtain the predicted coding rate of the color information of the dynamic point cloud, denoted as R c,cur , Then calculate the remaining bit rate, denoted as R res , R res =R target -R c,cur ; Then calculate the current predicted color coding quantization parameter QP c,cur The corresponding current prediction geometry coding quantization parameter is denoted as QP g,cur , Finally, according to step 3 Calculate the encoding quantization parameter set (QP g,cur ,QP c,cur ) represents the predicted total coding rate of the dynamic point cloud, denoted by R(QP g,cur ,QP c,cur ), Among them, QP c,cur ∈[QP c,start ,51],QP g,cur ∈[QP g,final ,51].

[0036] Step 9: Determine R(QP g,cur ,QPc,cur ) Whether the conditions are met If satisfied, the quantization parameter group (QP g,cur ,QP c,cur ) is added to the candidate list, and then step 10 is performed; otherwise, the encoding quantization parameter group (QP g,cur ,QP c,cur ), and then execute step 10; wherein the symbol “||” is an absolute value symbol, T represents the limit bit error rate, T∈[0,1], and T=0.03 ​​in this embodiment.

[0037] Step 10: Set the interval [QP c,start ,51] takes the next traversal value as the current value, and then returns to step 8 to continue until the interval [QP c,start ,51], each value in the list is traversed and a candidate list is obtained.

[0038] Step 11: According to S(QP in step 5 g ,QP c )=S g (QP g )+S c (QP c ), calculate the quantized subjective quality of the dynamic point cloud represented by each set of coding quantization parameter groups in the candidate list, and take the kth set of coding quantization parameter groups (QP g,k ,QP c,k The quantized subjective quality of the dynamic point cloud represented by ) is denoted as S(QP g,k ,QP c,k ), S(QP g,k ,QP c,k )=S g (QP g,k )+S c (QP c,k ), where 1≤k≤K, K represents the number of coding quantization parameter groups included in the candidate list, S g (QP g,k )pass Calculated, S c (QP c,k )pass Calculated.

[0039] Step 12: Find the quantized subjective quality with the largest value from the K quantized subjective qualities calculated in step 11; then use a set of coding quantization parameter groups corresponding to the quantized subjective quality with the largest value as a set of coding quantization parameter groups with the best subjective quality; and then encode the dynamic point cloud according to the set of coding quantization parameter groups with the best subjective quality.

[0040] In order to further illustrate the performance of the method of the present invention, the method of the present invention was tested.

[0041] To illustrate the performance of the method of the present invention, experiments were conducted in the V-PCC standard reference software TMC2 V9.0 and the corresponding HEVC reference software HM16.20-SCM8.8. The dynamic point cloud sequences of "Longdress", "Loot", "Queen", "Redandblack" and "Soldier" were selected as standard test sequences. The number of test frames of each standard test sequence was 32 frames, the encoding configuration was low latency, and the length of the picture group (GOP) was 16. Different target encoding rates were set according to the complexity of each standard test sequence. The four groups of target encoding rates for "Longdress" and "Redandblack" were 85kbpmp, 280kbpmp, 385kbpmp and 675kbpmp respectively; the four groups of target encoding rates for "Loot", "Queen" and "Soldier" were 46kbpmp, 56kbpmp, 72kbpmp and 120kbpmp respectively. The comparison method used in the test is Liu Q, Yuan H, Hou J, et al. Model-based joint bit allocation between geometry and color for video-based 3D point cloud compression [J]. IEEE Transactions on Multimedia, 2020, 23: 3278-3291. (Model-based joint bit allocation between geometry and color for video-based 3D point cloud compression, IEEE Transactions on Multimedia, 2020, 23: 3278-3291).

[0042] Table 1 lists the specific information of the dynamic point cloud sequences of “Longdress”, “Loot”, “Queen”, “Redandblack” and “Soldier”.

[0043] Table 1 Related parameter information of each dynamic point cloud sequence

[0044] Dynamic point cloud sequence Frame rate Frame rate Points Geometric accuracy Color Accuracy Longdress 300 30 ～800,000 10bit 8bit Loot 300 30 ～780,000 10bit 8bit Queen 250 50 ～1,000,000 10bit 8bit Redandblack 300 30 ～700,000 10bit 8bit Soldier 300 30 ～1,500,000 10bit 8bit

[0045] Table 2 lists the comparison of the coding rate allocation accuracy of the dynamic point cloud sequence listed in Table 1 using the method of the present invention and the comparative method. The coding rate allocation accuracy is reflected in two aspects: bit error rate and quantization parameter error. The bit error rate (BE) is defined as The symbol “||” is the absolute value symbol, R act Indicates the actual encoding rate, R targetrepresents the target encoding bit rate. The quantization parameter error (QP error, QPE) is defined as QPE = |QP g,B -QP g,A |+|QP c,B -QP c ,A|, where (QP g,B ,QP c,B ) is a set of coding quantization parameter groups (ideal state) with the minimum bit error rate (i.e., the optimal bit error rate) in the candidate list, (QP g,A ,QP c,A ) is a set of encoding quantization parameter groups selected for actual encoding.

[0046] Table 2 Comparison of coding rate allocation accuracy of the dynamic point cloud sequence listed in Table 1 using the method of the present invention and the comparative method

[0047]

[0048] In Table 2, the optimal bit error rate refers to the selection of (QP g,B ,QP c,B ) is the bit error rate of the encoding result. The actual bit error rate refers to the bit error rate of (QP g,A ,QP c,A ) of the coding result. It can be seen from Table 2 that the average optimal bit error rates of the method of the present invention and the comparative method are 1.95% and 2.25% respectively, and the actual average bit error rates of the method of the present invention and the comparative method are 3.17% and 4.1% respectively. From the results, since the preprocessing process (i.e., the process of fitting parameters) of the method of the present invention is slightly more complicated than that of the comparative method, the average optimal bit error rate can be regarded as similar, but the actual average bit error rate is reduced by 0.93%, and the actual average bit error rate is significantly reduced. It can be seen from Table 2 that the average QPE of the method of the present invention is 0.8 higher than that of the comparative method, indicating that the set of coding quantization parameter groups selected by the comparative method for actual coding is closer to the set of coding quantization parameter groups with the smallest bit error rate. However, when the QPE of the method of the present invention is larger, its bit error rate is lower, which also indicates that the coding quantization parameters obtained by the method of the present invention are more accurate, and the geometric coding quantization step size Q selected in the preprocessing process (i.e., the process of fitting parameters) is more accurate. g and color-coded quantization step size Q c More reasonable.

[0049] Table 3 lists the subjective performance of the method of the present invention for encoding the dynamic point cloud sequences listed in Table 1. For the subjective, the adjectival categorical judgment (ACJ) method is used for subjective scoring, wherein a negative score indicates that the subjective quality of encoding using the method of the present invention is worse than that of the original platform, and a positive score indicates that the subjective quality of encoding using the method of the present invention is better than that of the original platform.

[0050] Table 3 Subjective performance of encoding the dynamic point cloud sequence listed in Table 1 using the method of the present invention

[0051] Dynamic point cloud sequence 1 2 3 4 Longdress 0.27 0.47 0.07 0.07 Loot 0.00 0.27 0.20 0.73 Queen 0.20 0.13 0.60 0.40 Redandblack -0.13 0.00 0.07 0.00 Soldier -0.07 0.13 0.33 0.20 average 0.05 0.20 0.25 0.28

[0052] In Table 3, 1 to 4 correspond to the four target encoding bit rates of "Longdress" and "Redandblack", which are 85kbpmp, 280kbpmp, 385kbpmp and 675kbpmp respectively, and the four target encoding bit rates of "Loot", "Queen" and "Soldier" are 46kbpmp, 56kbpmp, 72kbpmp and 120kbpmp respectively.

[0053] It can be seen from Table 3 that as the target coding bit rate increases, the subjective quality of the reconstructed point cloud using the method of the present invention is better than that using the comparison method. This is because when the target coding bit rate is large, the geometric coding quantization parameter in the method of the present invention is closer to that of the comparison method, and the degree of geometric coding distortion is similar. However, the method of the present invention also ensures the subjective quality of the color of the reconstructed point cloud. Therefore, the subjective performance of the reconstructed point cloud using the method of the present invention is better.

Claims

1. A dynamic point cloud coding rate allocation method based on perceptual analysis, characterized in that The following steps are involved: Step 1: Under the V-PCC standard, the encoding rate of the geometric information of the dynamic point cloud is recorded as R g , and described as: The encoding rate of the color information of the dynamic point cloud is recorded as R c , and described as: Among them, α g and β g is the geometric encoding parameter, Q g represents the geometric coding quantization step size, α c and β c is the color coding parameter, Q c Indicates the color coding quantization step size; Step 2: Quantize step size Q in geometric coding g The geometric information of the dynamic point cloud is encoded when the values ​​are 3.25, 6.5, 16, and 40, respectively, corresponding to the actual encoding bit rate of the geometric information of the dynamic point cloud; then, according to the geometric coding quantization step size Q g The four values ​​of and the corresponding four actual encoding rates are Perform the least squares fitting and get α g and β g The value of Similarly, the color coding quantization step size Q c The color information of the dynamic point cloud is encoded when the values ​​are 16, 32, and 64 respectively, corresponding to the actual encoding bit rate of the color information of the dynamic point cloud; then according to the color coding quantization step size Q c The three values ​​of and the corresponding three actual encoding rates are Perform the least squares fitting and get α c and β c The value of Step 3: Based on step 2, and Get the encoding quantization step size group (Q g ,Q c ) represents the total encoding rate of the dynamic point cloud, denoted by R(Q g ,Q c ) and described as: Then The geometric coding quantization step size Q in g Convert to geometric coding quantization parameter QP g , color-coded quantization step size Q c Convert to color encoding quantization parameter QP c , get the encoding quantization parameter group (QP g ,QP c ) represents the total encoding bit rate of the dynamic point cloud, denoted as R(QP g ,QP c ) and described as: Wherein, f() represents the process of converting the coding quantization parameter to the coding quantization step size; Step 4: Geometry-based coding quantization parameter QP g A quantitative subjective quality model of the geometric information of dynamic point clouds is established, which is described as: And based on the color encoding quantization parameter QP c A quantitative subjective quality model of the color information of dynamic point clouds is established, which is described as: Among them, S g (QP g ) represents the quantization parameter QP based on geometric coding g Quantifying the subjective quality of the geometric information of a dynamic point cloud, S m Indicates the highest quantized subjective quality set, QP g,v Indicates the set geometric coding quantization parameter threshold, S c (QP c ) represents the color-coded quantization parameter QP c The quantified subjective quality of the color information of the dynamic point cloud, QP c,v Indicates the set color-coded quantization parameter threshold; Step 5: Based on step 4, and Get the encoding quantization parameter group (QP g ,QP c ) is represented by the quantized subjective quality model of the dynamic point cloud, which is described as: S(QP g ,QP c )=S g (QP g )+S c (QP c ); where S(QP g ,QP c ) indicates the encoding quantization parameter group (QP g ,QP c ) represents the quantified subjective quality of the dynamic point cloud; Step 6: Determine the geometric coding quantization parameter QP g The termination value is denoted as QP g,final , And determine the color encoding quantization parameter QP c The starting value is denoted as QP c,start , Among them, f -1 () represents the process of converting the coding quantization step size to the coding quantization parameter, R target represents the target encoding bit rate to be allocated; Step 7: Traverse the interval [QP c,start ,51], define the currently traversed value as the current value; Step 8: Use the current value as the current predicted color encoding quantization parameter and record it as QP c,cur ; Then according to QP c,cur Encode the color information of the dynamic point cloud to obtain the predicted coding rate of the color information of the dynamic point cloud, denoted as R c,cur , Then calculate the remaining bit rate, denoted as R res , R res =R target -R c,cur ; Then calculate the current predicted color coding quantization parameter QP c,cur The corresponding current prediction geometry coding quantization parameter is denoted as QP g,cur , Finally, according to step 3 Calculate the encoding quantization parameter set (QP g,cur ,QP c,cur ) represents the predicted total coding rate of the dynamic point cloud, denoted by R(QP g,cur ,QP c,cur ), Among them, QP c,cur ∈[QP c,start ,51],QP g,cur ∈[QP g,final ,51]; Step 9: Determine R(QP g,cur ,QP c,cur ) Whether the conditions are met If satisfied, the quantization parameter group (QP g,cur ,QP c,cur ) is added to the candidate list, and then step 10 is performed; otherwise, the encoding quantization parameter group (QP g,cur ,QP c,cur ), and then execute step 10; wherein the symbol "||" is an absolute value symbol, T represents the limit bit error rate, T∈[0,1]; Step 10: Set the interval [QP c,start ,51] takes the next traversal value as the current value, and then returns to step 8 to continue until the interval [QP c,start ,51], each value in the traversal is completed, and the candidate list is obtained; Step 11: According to S(QP in step 5 g ,QP c )=S g (QP g )+S c (QP c ), calculate the quantized subjective quality of the dynamic point cloud represented by each set of coding quantization parameter groups in the candidate list, and take the kth set of coding quantization parameter groups (QP g,k ,QP c,k The quantized subjective quality of the dynamic point cloud represented by ) is denoted as S(QP g,k ,QP c,k ), S(QP g,k ,QP c,k )=S g (QP g,k )+S c (QP c,k ), where 1≤k≤K, K represents the number of coding quantization parameter groups included in the candidate list, S g (QP g,k )pass Calculated, S c (QP c,k )pass Calculated; Step 12: Find the quantized subjective quality with the largest value from the K quantized subjective qualities calculated in step 11; then use a set of coding quantization parameter groups corresponding to the quantized subjective quality with the largest value as a set of coding quantization parameter groups with the best subjective quality; and then encode the dynamic point cloud according to the set of coding quantization parameter groups with the best subjective quality.

2. According to claim 1, a method for allocating bit rate of dynamic point cloud coding based on perceptual analysis is characterized in that In step 4, S m =51,QP g,v =10,QP c,v =28.

3. According to the method of dynamic point cloud coding rate allocation based on perceptual analysis in claim 1, it is characterized in that In the step 9, T=0.03.

Images