CN103997646B - Fast intra-mode prediction mode selecting method in a kind of HD video coding - Google Patents
Fast intra-mode prediction mode selecting method in a kind of HD video coding Download PDFInfo
- Publication number
- CN103997646B CN103997646B CN201410200667.6A CN201410200667A CN103997646B CN 103997646 B CN103997646 B CN 103997646B CN 201410200667 A CN201410200667 A CN 201410200667A CN 103997646 B CN103997646 B CN 103997646B
- Authority
- CN
- China
- Prior art keywords
- prediction
- mode
- prediction modes
- prediction mode
- coding
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The present invention proposes fast intra-mode prediction mode selecting method in a kind of HD video coding.Residual error is handled first with absolute error and SAD, selects the small several candidate modes of error.Then using simplified rate distortion costs function, that is the cost function of Hadamard transform absolute value and SATD (Sum of Absolute Hadamard Transformed Difference) further reduces candidate modes number, and candidate modes are obtained by the way of threshold value, the candidate modes number of the predicting unit of formed objects may also differ.Followed by the correlation of information of forecasting between the probability statistics and neighboring prediction unit of optimization model, detect whether most probable prediction mode MPM (Most Probable Mode) is included in candidate modes, MPM refers to the optimum prediction mode of left side predicting unit encoded around current prediction unit and top predicting unit.If comprising only will be contained in the MPM in candidate modes as final RDO candidate modes；Otherwise directly using the candidate modes further screened as final RDO candidate modes.The optimal intra prediction mode of present encoding block is finally obtained according to Lagrangian rate-distortion optimization criterion.
Description
The technical field is as follows:
the invention belongs to the field of video compression coding in signal processing, and particularly provides a rapid intra-frame prediction mode selection method aiming at the field of new generation video coding, which accelerates the intra-frame prediction coding speed of high definition video on the premise of ensuring the image quality.
Background art:
high Efficiency Video Coding (HEVC) is a new generation Video Coding standard jointly established by ISO/IEC MPEG and ITU-T VCEG, inherits part of the technology of the existing Video Coding standard H.264/AVC, improves the related technology at the same time, and aims to double the compression Efficiency on the basis of the original Video decoding quality. That is, the encoding Complexity can be properly increased to ensure that the video image quality is not changed, and the code stream of the video is reduced by 50% (see m.viitanen, j.vanne, t.d.hamalanen, et al.
HEVC still adopts a block-based coding framework, and simultaneously adopts a new image representation method, i.e., a larger partition block is used to code an image, thereby obtaining better coding performance (see g.j.subllivan, j.r.ohm.recording definitions in a decoding of High Efficiency Video Coding (HEVC) [ a)].International Society for Optics and Photonics[C]San Diego, CA, united states,2010, 7798. A Largest Coding Unit (LCU), a Coding Unit (CU), a Prediction Unit (PU), a Transform Unit (TU) are adopted as basic processing units. A Coding Unit (CU) is a basic unit divided at the time of coding, a Prediction Unit (PU) is a basic unit used at the time of prediction, and a Transform Unit (TU) is a basic unit used for transform quantization. The flexible block structure division mode has the division depth up to 4, namely CU 0 Size of 64 × 64 (depth of 0), CU 1 Size 32 × 32 (depth 1), CU 2 Size 16 × 16 (depth 2), CU 3 The size is 8 x 8 (depth is 3), and 4 coding blocks of different sizes participate in the selection calculation of the optimal block. The block division structure can optimize the operation of coding prediction, transformation and the like of each block. Meanwhile, HEVC increases the number of intra-frame prediction modes, so that each PU has 33 prediction angles, and DC prediction and plane prediction are added to obtain 35 intra-frame prediction modes, thereby improving the video coding efficiency.
However, these improvements also inevitably increase the coding complexity while improving the coding efficiency, and HEVC performs a Rate Distortion Optimization (RDO) process in all prediction modes using PUs of each size, so that the final rate distortion cost calculation number of one maximum coding unit is 7327, which makes the encoder operation very complex. Therefore, many people start to proceed with the optimization of the algorithm while the standard is continuously perfected. Because of the huge computation amount of intra-frame prediction, reducing the complexity of intra-frame prediction coding becomes one of the important research points for improving the coding efficiency. The complexity of an optimized intra-frame mode selection algorithm in H.264/AVC has achieved more research results, and the main methods thereof are divided into three categories: reducing the number of prediction modes through a certain rule; the number of prediction modes is not reduced, and the algorithm is optimized; and optimizing the structure of the software and hardware algorithm. Su Rui, etc. uses transform domain information to obtain the image texture direction, and reduces the number of prediction modes by the texture direction to realize fast intra-frame prediction coding (see Su Rui, liu Guizhong, zhang Tongyu, etc. the new algorithm [ J ] for fast realizing H.264 intra-frame prediction coding by using transform domain information, 2007,29 (1): 161-165.). Kim et al effectively reduces the encoding time of intra prediction using post-transform residual sum and gradient and reducing the number of candidate prediction modes (C Kim, H shift, C J kuo. Multistage mode decision for intra prediction in h.264codec [ C ]. Visual Communications and Image Processing,2004, 5308. And the research results of HEVC intra-frame prediction optimization algorithms at home and abroad are less. Zhao et al, which adopts a fast partitioning algorithm that does not reduce the number of prediction modes and obtains coding units according to coding complexity analysis of different coding units, to reduce complexity of prediction coding (see l.zhao, li Zhang, siwei Ma, et al, fast mode resolution algorithm for intra prediction in HEVC [ a ]. IEEE Visual Communications and Image Processing (VCIP) [ C ], tainan, taiwan,2011,1-4.). G.Van Wallendael and the like compare rate distortion costs after Hadamard transformation, roughly select a certain number of prediction modes with low residual cost, and then perform rate distortion optimization on the modes to obtain a better effect, but the prediction time occupies a longer time and needs to be further optimized (see G.Van Wallendael, S.Van Leuven, J.De Cock, et al. Improved intra mode signaling for HEVC [ A ], IEEE International Conference on Multimedia and Expo [ C ], barcelona, spain,2011,1-6.).
In order to solve the problem of high complexity of intra-frame prediction coding in the current high-definition video coding, the method fully utilizes probability statistics of an optimal mode and the correlation of prediction information between adjacent prediction units, reduces the number of intra-frame prediction modes by adopting Absolute error Sum (SAD) and simplified rate-distortion cost models, and finally selects the optimal intra-frame prediction mode from less candidate modes by Lagrange rate-distortion optimization so as to achieve the purpose of quick intra-frame prediction mode selection in the high-definition video coding. Experiments prove that the method greatly reduces the intra-frame prediction complexity and obviously reduces the intra-frame prediction time on the premise of ensuring the high-definition video coding efficiency.
The invention content is as follows:
the invention provides a method for selecting a fast intra-frame prediction mode in high-definition video coding. The residual is first processed using the absolute error and SAD to select several candidate prediction modes with small errors. Then, the number of candidate prediction modes is further reduced by using a simplified rate-distortion cost function, namely a cost function of a Hadamard transform (Hadamard transform) Absolute value and a cost function of a SATD (Sum of Absolute Transformed Difference), and the candidate prediction modes are obtained by using a threshold method, and the number of candidate prediction modes of prediction units with the same size may be different. Then, it is detected whether a Most Probable prediction Mode MPM (best predicted Mode) is included in the candidate prediction modes, wherein the MPM refers to the best prediction Mode of a left prediction unit and an upper prediction unit that have been coded around the current prediction unit, by using the probability statistics of the best prediction Mode and the correlation of prediction information between neighboring prediction units. If so, only the MPMs included in the candidate prediction mode are taken as final RDO candidate prediction modes; otherwise, directly taking the candidate prediction mode further screened as the final RDO candidate prediction mode. And finally, obtaining the optimal intra-frame prediction mode of the current coding block according to the Lagrange rate distortion optimization criterion.
A method for selecting a fast intra-frame prediction mode in high-definition video coding is characterized by comprising the following implementation steps:
the method comprises the following steps: respectively calculating absolute values and SADs of prediction residuals of 35 intra-frame prediction modes of a coding unit with the size of 64 multiplied by 64 at present, sorting according to the sizes, and selecting a plurality of smaller candidate prediction modes;
step two: calculating candidate prediction modes obtained after screening in the first step by using simplified rate-distortion cost, namely a cost function of a Hadamard transform absolute value and SATD (SATD), selecting a plurality of prediction modes with smaller simplified rate-distortion cost, and reserving the number of different candidate prediction modes in a threshold value mode according to the statistical probability that the prediction modes become the final optimal intra-frame prediction mode;
step three: detecting whether the MPM (most probable prediction mode) is contained in the candidate prediction modes, and if so, taking the MPM contained in the candidate prediction modes as a final RDO candidate prediction mode; otherwise, taking the candidate prediction mode obtained in the second step as a final RDO candidate prediction mode, wherein MPM refers to the best prediction mode of the left prediction unit and the upper prediction unit which are coded around the current prediction unit;
step four: calculating rate distortion costs RDO of all prediction modes in the RDO candidate prediction modes, and selecting the prediction mode with the minimum rate distortion cost as the optimal prediction mode; if only one candidate mode exists in the RDO candidate prediction modes, directly taking the candidate mode as the optimal prediction mode of the prediction unit; obtaining a rate distortion cost RdCost1 of the optimal prediction mode;
step five: repeating the first step to the fourth step on the four sub-coding units obtained after the coding unit is divided to select the optimal prediction mode for coding, and obtaining the sum RdCost2 of the rate-distortion cost of the four divided sub-coding units;
step six: comparing the rate-distortion cost of the coding unit and the rate-distortion cost of the sub-coding unit, namely comparing the values of CostRd1 and CostRd2, and when RdCost1> RdCost2, turning to step seven; otherwise, the current coding unit is adopted for predictive coding, and the step eight is carried out;
step seven: dividing the sub-coding unit into four smaller coding units, repeating the fifth step and the sixth step, and if the size of the coding unit is 4 multiplied by 4, turning to the eighth step;
step eight: the intra prediction encoding of the current encoding unit ends.
According to the method for selecting the fast intra-frame prediction mode in the high-definition video coding, the processed video sequence is in a YUV format, and the intra-frame prediction of the brightness component is processed by adopting the eight steps.
The simplified rate distortion cost in the second step, namely the cost function of the absolute value of the Hadamard transform and the SATD is as follows:
J SATD ＝SATD+λ mode ·B mode (1)
wherein, SATD is the sum of absolute values of residual errors obtained after Hadamard transform is performed on the prediction residual errors, namely:
λ mode is the lagrange factor and is calculated using the following formula:
λ mode ＝0.57×2 ((QP-12)/3.0) (3)
B mode is the number of bits required by the recorded intra mode in the transmission of the code stream.
The probability that the candidate prediction modes finally become the optimal modes of the prediction unit is not necessarily equally distributed after the candidate prediction modes are screened by the SATD rate-distortion cost criterion in the step two, and the probability that the prediction mode arranged in the front is the optimal prediction mode is far greater than the probability that the prediction mode arranged in the back. Meanwhile, the number of the prediction modes of different prediction units is different, so the probability that the two prediction modes which are screened before are the optimal prediction modes is also different. In the invention, the candidate prediction modes are obtained by adopting a threshold value mode, and the number of the candidate prediction modes of the PUs with the same size can be different. When the simplified rate-distortion cost of a prediction mode satisfies the following condition, then this prediction mode is considered as an RDO candidate prediction mode.
In the formula (4), the reaction mixture is,is a reduced rate distortion of the current prediction modeThe cost is that the cost of the product,the minimum value of the simplified rate-distortion cost of all the prediction modes of the current prediction unit is T, T is a threshold value, T =1.2 is taken in the invention, and the prediction result of the final prediction unit is the best at the moment.
And in the fourth step, rate distortion costs of all the prediction modes in the RDO candidate prediction modes are calculated, wherein the cost function is as follows:
J(o,r,MODE|QP,λ n )＝SSD(o,r,MODE|QP)+λ n R(o,r,MODE|QP) (5)
wherein QP is the quantization parameter of the quantization unit, λ n Is the lagrange multiplier, related to QP, SSD (-) is the sum of the squared differences between the original luma coding unit o and the reconstructed coding unit R, R (-) is the number of bits coded with MODE.
In the formula, orig L (x, y) is the pixel value of the original luminance coding unit, rec L (x, y) are the pixel values of the reconstructed luma coding unit.
The method for selecting the fast intra-frame prediction mode in the high-definition video coding has the advantages that:
(1) The method first roughly selects a prediction mode for the first time according to the Sum of Absolute Differences (SAD) of prediction residuals so as to reduce the number of candidate prediction modes.
(2) The method adopts simplified rate distortion cost functions, namely Hadamard transform absolute values and SATD cost functions as a transform domain characteristic to carry out secondary screening on the prediction mode after the first coarse selection. Compared with SAD, the SATD size is more closely related to the distortion size, namely, the RDO performance is more relevant; meanwhile, the algorithm of the Hadamard transform can be realized by adopting addition and shift operation, and the calculation complexity is reduced. According to the probability statistics of the optimal mode, for different prediction units, a threshold value method is adopted to obtain candidate prediction modes, and the number of the candidate prediction modes of PUs with the same size may also be different.
(3) The method utilizes the correlation of the prediction modes of adjacent prediction units and the high probability that the most probable prediction mode MPM (the best prediction mode of a left prediction unit and an upper prediction unit which are coded around the current prediction unit) becomes the final optimal prediction mode, detects whether the MPM is contained in the candidate prediction modes after rough selection, directly takes the MPM as an RDO candidate prediction mode for the containing condition, carries out rate distortion cost calculation, and selects the optimal prediction mode from the MPM.
Description of the drawings:
FIG. 1 is a flowchart illustrating a fast intra prediction mode selection method in high definition video coding according to the present invention;
FIG. 2 is a diagram showing the prediction directions of 33 angular prediction modes in a first step of the fast intra prediction mode selection method for high definition video coding according to the present invention;
FIG. 3 is a flowchart illustrating the calculation of the rate-distortion cost in the fourth step of the method for selecting a fast intra-frame prediction mode in high definition video coding according to the present invention;
FIG. 4 is a graph of PSNR comparison of the peak signal-to-noise ratio for the first 30 frames of a Cactus video sequence, respectively, using the method of the present invention and HM4.0 compression coding;
FIG. 5 is a plot of the code rate BR for the first 30 frames of the Captus video sequence respectively encoded using the method of the present invention and HM4.0 compression;
FIG. 6 is a graph comparing rate-distortion curves for the first 30 frames of a BQTerace video sequence, respectively, encoded using the method of the present invention and HM4.0 compression;
FIG. 7 is a graph comparing the average encoding time at different code rates (i.e., different values for QP) for the first 30 frames of standard high definition video sequences BQTerace, cattus, chinasSpeed, flowervase and FourPeople using HM4.0 and the method of the present invention;
FIG. 8 (a) is the 4 th frame original image of the high definition video sequence BQTerace;
FIG. 8 (b) is the result image of encoding and decoding the 4 th frame of the high definition video sequence BQTerace by using HM 4.0;
FIG. 8 (c) is the result image of the high definition video sequence BQTerace frame 4 after being encoded and decoded by the method of the present invention.
The specific implementation mode is as follows:
the method of the present invention will be described in further detail with reference to the accompanying drawings.
The invention provides a method for selecting a fast intra-frame prediction mode in high-definition video coding. The residual is first processed using the absolute error and SAD to select several candidate prediction modes with small errors. Then, the number of candidate prediction modes is further reduced by using a simplified rate-distortion cost function, namely, a cost function of a Hadamard transform (Hadamard transform) Absolute value and a cost function of a SATD (Sum of Absolute Transformed) function, and the candidate prediction modes are obtained by using a threshold method, and the number of the candidate prediction modes of prediction units with the same size may be different. Then, using the probability statistics of the optimal Mode and the correlation of prediction information between adjacent prediction units, it is detected whether a Most Probable prediction Mode MPM (Most Probable Mode) is included in the candidate prediction modes, where the MPM refers to the optimal prediction Mode of the left prediction unit and the upper prediction unit that have been encoded around the current prediction unit. If so, only the MPMs contained in the candidate prediction modes are taken as final RDO candidate prediction modes; otherwise, directly taking the candidate prediction mode further screened as the final RDO candidate prediction mode. And finally, obtaining the optimal intra-frame prediction mode of the current coding block according to the Lagrange rate distortion optimization criterion.
As shown in fig. 1, a method for selecting a fast intra prediction mode in high definition video coding includes the following steps:
the method comprises the following steps: the absolute values and SADs of prediction residuals of 35 intra prediction modes of a coding unit with the size of 64 x 64 at present are respectively calculated, sorting is carried out according to the sizes, and a plurality of smaller candidate prediction modes are selected. The 35 intra prediction modes include 33 angular prediction modes, a DC prediction mode, and a planar prediction mode. The prediction directions of the 33 angle prediction modes are shown in figure 2.
Calculating the difference value of corresponding pixels between the original image coding unit and the prediction coding unit:
Diff(x,y)＝Orig(x,y)-Pred(x,y) (7)
in the formula, origin (x, y) and Pred (x, y) are the encoding unit original pixel value and the predicted pixel value, respectively. Diff (x, y) is the residual of the original and predicted pixel values.
Step two: and calculating candidate prediction modes obtained after the step one screening by adopting simplified rate-distortion cost, namely a cost function of a Hadamard transform absolute value and SATD, selecting a plurality of prediction modes with lower simplified rate-distortion cost, and reserving the number of different candidate prediction modes in a threshold value mode according to the statistical probability that the prediction modes become the final optimal intra-frame prediction mode.
The simplified rate-distortion cost, i.e., the cost function of the hadamard transform absolute values and the SATD, is:
J SATD ＝SATD+λ mode ·B mode (9)
wherein, SATD is the sum of absolute values of residual errors obtained after Hadamard transform is performed on the prediction residual errors, namely:
λ mode is the lagrangian factor and is calculated using the following formula:
λ mode ＝0.57×2 ((QP-12)/3.0) (11)
B mode is the number of bits required by the recorded intra mode in the transmission of the code stream.
The probability of the candidate prediction mode finally becoming the optimal mode of the prediction unit is not necessarily equally distributed through screening of the SATD rate-distortion cost criterion, and the probability of the prediction mode arranged in the front in the criterion becoming the optimal prediction mode is far greater than that of the prediction mode arranged in the back. Meanwhile, the number of the prediction modes of different prediction units is different, so the probability that the two prediction modes which are screened before are the optimal prediction modes is also different. In the invention, the candidate prediction modes are obtained by adopting a threshold value mode, and the number of the candidate prediction modes of the PUs with the same size can be different. When the simplified rate-distortion cost of a prediction mode satisfies the following condition, then this prediction mode is considered as an RDO candidate prediction mode.
In the formula (12), the reaction mixture is,is the reduced rate-distortion cost of the current prediction mode,the prediction method is the minimum value of the simplified rate distortion cost of all the prediction modes of the current prediction unit, T is a threshold, T =1.2 is taken in the invention, and the prediction result of the final prediction unit is the best at the moment.
Step three: detecting whether the MPM (most probable prediction mode) is contained in the candidate prediction modes, and if so, taking the MPM contained in the candidate prediction modes as a final RDO candidate prediction mode; otherwise, directly taking the candidate prediction mode obtained in the step two as the final RDO candidate prediction mode.
MPM refers to the best prediction mode of the left prediction unit and the upper prediction unit that have been encoded around the current prediction unit, and is very important in mode selection. Table 1 is a statistical analysis of the proportion of the most probable prediction mode MPM among all the optimal prediction modes. As shown in table 1, for different resolution video sequences, MPM accounts for a high percentage of all the best prediction modes. Therefore, MPM has a higher probability of becoming the optimal prediction mode.
TABLE 1 proportion of MPMs in all optimal prediction modes
Step four: calculating rate distortion costs of all prediction modes in the RDO candidate prediction modes, and selecting the prediction mode with the minimum rate distortion cost as the optimal prediction mode; if only one candidate mode exists in the RDO candidate prediction modes, directly taking the candidate mode as the optimal prediction mode of the prediction unit; the rate-distortion cost rdcmos 1 of the best prediction mode is obtained.
The calculation flow chart of the rate distortion cost is shown in figure 3. The calculation formula is as follows:
J(o,r,MODE|QP,λ n )＝SSD(o,r,MODE|QP)+λ n R(o,r,MODE|QP) (13)
wherein QP is the quantization parameter of the quantization unit, λ n Is the lagrange multiplier, related to QP, SSD (-) is the sum of the squared differences between the original luma coding unit o and the reconstructed coding unit R, R (-) is the number of bits coded with MODE.
In the formula, orig L (x, y) is the pixel value of the original luminance coding unit, rec L (x, y) are the pixel values of the reconstructed luma coding unit.
Step five: repeating the first step to the fourth step on the four sub-coding units obtained after the coding unit is divided to select the optimal prediction mode for coding, and obtaining the sum RdCost2 of the rate-distortion cost of the four divided sub-coding units;
step six: comparing the rate-distortion cost of the coding unit and the rate-distortion cost of the sub-coding unit, namely comparing the values of CostRd1 and CostRd2, and when RdCost1> RdCost2, turning to step seven; otherwise, the current coding unit is adopted for predictive coding, and the step eight is carried out;
step seven: dividing the sub-coding unit into four smaller coding units, repeating the fifth step and the sixth step, and if the size of the coding unit is 4 multiplied by 4, turning to the eighth step;
step eight: the intra prediction encoding of the current encoding unit ends.
The video sequence processed by the method is in YUV format, the method is verified on the basis of the HM4.0 code of HEVC reference software of a new generation of video coding standard, and the used computer isThe memory physical address expansion of core TM2i3-2120CPU,3.30GHz,4GB, the experimental environment is Visual Studio2008. Simulation experiments were performed on the first 30 frames of the video sequences BQTerrace, cactus, chinampeed, flowervase, and FourPeople at different resolutions. The experimental parameter configuration is shown in table 2.
TABLE 2 Experimental parameter configuration
| Software | HM4.0 |
| Quantization parameter QP | 27,30,32,34,36 |
| Number of encoded frames | 30 |
| Frame rate | 60Hz |
| GOP structure | IIII… |
| Entropy coding | CABAC |
Because the method of the present invention mainly aims at the problem of the high complexity of a new generation video coding system, the complexity of the intra-frame prediction coding is represented by the number of clock cycles required by the intra-frame prediction coding, and specific quantitative analysis on the calculation complexity of the method of the present invention and the standard implementation algorithm in the HM4.0 code is shown in table 3. In the experiment, the statistical experiment is carried out on the videos with different resolutions, and QP =32 is adopted for each video, so that the method greatly accelerates the intra-frame prediction encoding speed, the intra-frame prediction complexity is reduced by 71.49% in the highest degree and is reduced by 49.82% on average compared with the algorithm complexity before optimization.
Table 4 shows the performance comparison between the method of the present invention and the intra-frame predictive coding algorithm in the HM4.0 standard test model in terms of peak signal-to-noise ratio PSNR, coding rate BR, and coding Time. It can be seen from the table that the coding efficiency of the video coding system of the method of the present invention is slightly reduced compared with that of HM4.0, but the intra-frame coding complexity is greatly reduced, and the average is reduced by 30.82%. Δ PSNR, Δ Time, Δ BR in the table are defined as follows:
ΔPSNR＝PSNR OURS –PSNR HM4.0 (15)
ΔBR＝(BR OURS –BR HM4.0 )/BR HM4.0 (16)
ΔTime＝(Time OURS –Time HM4.0 )/Time HM4.0 (17)
TABLE 3 complexity comparison before and after intra prediction optimization
TABLE 4 comparison of coding Performance of the inventive method with the HM4.0 method
| Video sequence and resolution | ΔPSNR/dB | ΔBR/％ | ΔTime/％ |
| BQTerrace(1920×1080) | -0.15 | 2.34 | -30.75 |
| Cactus(1920×1080) | -0.1 | 2.52 | -28.58 |
| ChinaSpeed(1024×768) | -0.31 | -30.39 | 3.49 |
| FourPeople(1280×720) | -0.18 | -33.63 | 3.90 |
| Flowervase(832×480) | -0.22 | -30.74 | 2.41 |
| Mean value of | -0.19 | -30.82 | 2.94 |
For more intuitive description of the variation of the encoding performance, fig. 4-6 show the peak signal-to-noise ratio PSNR and the encoding rate BR of the video decoded by the HM4.0 and the method of the present invention for the first 30 frames of the high definition video sequence, respectively. As can be seen from the figure, the high definition video coding performance of the method of the invention has no obvious change compared with the coding performance of the test model HM 4.0.
Fig. 8 (a), 8 (b), and 8 (c) show the original image of the 4 th frame of the video sequence bqterace, the result image decoded by the method in HM4.0, and the result image decoded by the method of the present invention, respectively. It can be seen that the method of the present invention adopts the method of removing the prediction mode with the least accurate prediction priority to reduce the candidate prediction modes, which not only ensures the accuracy of the prediction, but also reduces the complexity of the prediction, and compared with the HM4.0 method, the subjective quality of the decoded image has almost no difference and is not much different from the subjective quality of the original image.
Therefore, under the condition that the influence of the coding efficiency is not large, compared with HM4.0, the intra-frame prediction time is averagely reduced by 30.82%, the intra-frame prediction coding complexity is reduced to a great extent, and the rapid selection of the intra-frame prediction mode of the high-definition video coding is realized.
Claims (3)
1. A method for selecting a fast intra-frame prediction mode in high-definition video coding is characterized by comprising the following steps:
the method comprises the following steps: respectively calculating absolute values and SADs (sum of absolute differences) of prediction residuals of 35 prediction modes of a coding unit with the size of 64 multiplied by 64 at present, sorting according to the sizes, and selecting a plurality of smaller candidate prediction modes;
step two: calculating candidate prediction modes obtained after screening in the first step by using simplified rate-distortion cost, namely a cost function of a Hadamard transform absolute value and SATD (saturated iterative reconstruction) to select a plurality of prediction modes with smaller simplified rate-distortion cost, and reserving RDO (remote desktop object) candidate prediction modes in a threshold value mode according to the statistical probability that the prediction modes become the final optimal intra-frame prediction modes; the probability of the candidate prediction mode obtained by screening simplified rate distortion cost, namely the Hadamard transform absolute value and the cost function of the SATD in the step two, which finally becomes the optimal mode of the prediction unit is not equally distributed, and the probability of the prediction mode arranged in the front becoming the optimal prediction mode is greater than the probability of the prediction mode arranged in the back; meanwhile, the number of the prediction modes of different prediction units is different, so the probability that the prediction modes of the two previous prediction modes after screening are the optimal prediction modes is also different, the candidate prediction modes are obtained by adopting a threshold value mode, when the simplified rate distortion cost of the prediction modes meets the following conditions, the prediction modes are considered as the RDO candidate prediction modes,(4)
in the formula (4), the reaction mixture is,is the reduced rate-distortion cost of the current prediction mode,the minimum value of the simplified rate-distortion cost of all prediction modes of the current prediction unit is T, T is a threshold value, and T =1.2 is taken;
step three: detecting whether the most probable prediction mode MPM is included in the candidate prediction modes in the step one, and if so, only using the MPM included in the candidate prediction modes as a final RDO candidate prediction mode; otherwise, taking the candidate prediction mode obtained in the step two as a final RDO candidate prediction mode;
step four: calculating rate distortion costs RDO of all prediction modes in the RDO candidate prediction modes, and selecting the prediction mode with the minimum RDO value as the optimal prediction mode; if only one candidate mode exists in the RDO candidate prediction modes, directly taking the candidate mode as the optimal prediction mode of the prediction unit; obtaining a rate distortion cost RdCost1 of the optimal prediction mode;
step five: respectively calculating absolute values and SADs (sum of absolute differences) of prediction residuals of 35 prediction modes of the four currently obtained sub-coding units by four sub-coding units obtained after the coding units with the size of 64 multiplied by 64 are divided, sorting according to the sizes, selecting a plurality of smaller candidate prediction modes, repeating the steps from the second step to the fourth step, selecting the optimal prediction mode for the sub-coding units, and coding to obtain the sum RdCost2 of rate distortion costs RDO of the four divided sub-coding units;
step six: comparing the optimal prediction mode with the rate-distortion cost RDO of the sub-coding unit, namely comparing the values of the RdCost1 and the RdCost2, and going to the step seven when the RdCost1> the RdCost2; otherwise, the coding unit with the current size of 64 multiplied by 64 is adopted to carry out predictive coding, and the step eight is carried out;
step seven: the sub-coding units are further divided into four smaller coding units, the absolute values and SADs of prediction residuals of 35 prediction modes of the four smaller coding units which are divided currently are calculated for the four smaller coding units which are further divided, sorting is carried out according to the sizes, several smaller candidate prediction modes are selected, the steps from the second step to the fourth step are repeated, the optimal prediction mode is selected for the sub-coding units to carry out coding, the step from the sixth step is repeated, and if the sizes of the four smaller coding units are 4 multiplied by 4, the step from the eighth step is carried out;
step eight: the current coding unit predictive coding ends.
2. The method of claim 1, wherein the intra prediction mode selection comprises: the processed video sequence is in YUV format, and the intra-frame prediction of the brightness component is processed by adopting the eight steps.
3. The method of claim 1, wherein the intra prediction mode selection comprises: the simplified rate distortion cost in the second step, namely the cost function of the absolute value of the Hadamard transform and the SATD is as follows:
J SATD ＝SATD+λ mode ·B mode (1)
wherein, SATD is the sum of absolute values of residual errors obtained after Hadamard transform of the prediction residual errors, i.e.:
wherein Diff (x, y) = origin (x, y) -Pred (x, y),origin (x, y) and Pred (x, y) are the original pixel value and the predicted pixel value of the coding unit, respectively, diff (x, y) is the residual between the original and predicted pixel values, diffHAD (x, y) represents the Hadamard transform of the predicted residual;
λ mode is the lagrangian factor and is calculated using the following formula:
λ mode ＝0.57×2 ((QP-12)3.0) (3)
wherein QP is a quantization parameter of a quantization unit;
B mode is the number of bits required by the recorded intra mode in the transmission of the code stream.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410200667.6A CN103997646B (en) | 2014-05-13 | 2014-05-13 | Fast intra-mode prediction mode selecting method in a kind of HD video coding |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410200667.6A CN103997646B (en) | 2014-05-13 | 2014-05-13 | Fast intra-mode prediction mode selecting method in a kind of HD video coding |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103997646A CN103997646A (en) | 2014-08-20 |
| CN103997646B true CN103997646B (en) | 2018-05-01 |
Family
ID=51311632
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410200667.6A Active CN103997646B (en) | 2014-05-13 | 2014-05-13 | Fast intra-mode prediction mode selecting method in a kind of HD video coding |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103997646B (en) |
Families Citing this family (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104301724B (en) | 2014-10-17 | 2017-12-01 | 华为技术有限公司 | Method for processing video frequency, encoding device and decoding device |
| CN104320660B (en) * | 2014-10-31 | 2017-10-31 | 中国科学技术大学 | Rate-distortion optimization method and coding method for lossless video encoding |
| CN104378643B (en) * | 2014-12-04 | 2017-11-14 | 南京理工大学 | A kind of 3D video depths image method for choosing frame inner forecast mode and system |
| CN104754338A (en) * | 2014-12-30 | 2015-07-01 | 北京奇艺世纪科技有限公司 | Selection method and device for intra-frame predication mode |
| US9883187B2 (en) * | 2015-03-06 | 2018-01-30 | Qualcomm Incorporated | Fast video encoding method with block partitioning |
| CN106375762B (en) * | 2015-07-22 | 2019-05-24 | 杭州海康威视数字技术股份有限公司 | Reference frame data compression method and its device |
| CN106488236B (en) * | 2015-09-01 | 2019-07-12 | 北京君正集成电路股份有限公司 | The calculation method and device of absolute transformed error sum in a kind of Video coding |
| CN105491390B (en) * | 2015-11-30 | 2018-09-11 | 哈尔滨工业大学 | Intra-frame prediction method in hybrid video coding standard |
| CN107306353B (en) * | 2016-04-19 | 2020-05-01 | 广州市动景计算机科技有限公司 | Image space prediction mode selection method and device, and image compression method and device |
| CN106534855B (en) * | 2016-11-04 | 2019-03-26 | 西安理工大学 | A kind of Lagrange factor calculation method towards SATD |
| CN108347602B (en) * | 2017-01-22 | 2021-07-30 | 上海澜至半导体有限公司 | Method and apparatus for lossless compression of video data |
| CN107222743B (en) * | 2017-07-04 | 2020-02-04 | 西安万像电子科技有限公司 | Image processing method, device and system |
| CN107371023A (en) * | 2017-08-07 | 2017-11-21 | 福建星网智慧科技股份有限公司 | The fast selecting method of luma prediction modes in a kind of frame |
| CN110049339B (en) * | 2018-01-16 | 2023-02-17 | 腾讯科技（深圳）有限公司 | Prediction direction selection method and device in image coding and storage medium |
| CN110166785B (en) * | 2018-07-25 | 2022-09-13 | 腾讯科技（深圳）有限公司 | Intra-frame prediction method and device, storage medium and electronic device |
| CN108881905B (en) * | 2018-08-04 | 2021-08-31 | 福州大学 | Probability-based intra-frame encoder optimization method |
| CN110198442B (en) * | 2018-12-27 | 2023-02-28 | 腾讯科技（深圳）有限公司 | Intra-frame prediction method, device and storage medium for video coding |
| CN109672895A (en) * | 2018-12-27 | 2019-04-23 | 北京佳讯飞鸿电气股份有限公司 | A kind of HEVC intra-frame prediction method and system |
| CN109889827B (en) * | 2019-04-11 | 2021-01-29 | 腾讯科技（深圳）有限公司 | Intra-frame prediction coding method and device, electronic equipment and computer storage medium |
| CN110213586B (en) * | 2019-06-10 | 2021-03-05 | 杭州电子科技大学 | VVC intra-frame prediction angle mode rapid selection method |
| CN112118444B (en) * | 2019-06-20 | 2022-11-25 | 杭州海康威视数字技术股份有限公司 | Encoding method and device |
| CN110896481B (en) * | 2019-12-31 | 2022-03-08 | 杭州当虹科技股份有限公司 | Fast intra mode coding method suitable for HEVC |
| CN113170144A (en) * | 2020-02-28 | 2021-07-23 | 深圳市大疆创新科技有限公司 | Video processing method and device |
| CN111294591B (en) * | 2020-05-09 | 2020-09-25 | 腾讯科技（深圳）有限公司 | Video information processing method, multimedia information processing method and device |
| CN111918059B (en) * | 2020-07-02 | 2023-01-13 | 北京大学深圳研究生院 | Hardware-friendly regression tree-based intra-frame prediction mode decision method and device |
| CN117156133B (en) * | 2023-10-27 | 2024-02-09 | 淘宝（中国）软件有限公司 | Inter-frame prediction mode selection method and device for video coding |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102665079A (en) * | 2012-05-08 | 2012-09-12 | 北方工业大学 | Adaptive fast intra prediction mode decision for high efficiency video coding (HEVC) |
| CN102932642A (en) * | 2012-11-13 | 2013-02-13 | 北京大学 | Interframe coding quick mode selection method |
-
2014
- 2014-05-13 CN CN201410200667.6A patent/CN103997646B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102665079A (en) * | 2012-05-08 | 2012-09-12 | 北方工业大学 | Adaptive fast intra prediction mode decision for high efficiency video coding (HEVC) |
| CN102932642A (en) * | 2012-11-13 | 2013-02-13 | 北京大学 | Interframe coding quick mode selection method |
Non-Patent Citations (2)
| Title |
|---|
| HEVC帧内模式决策及其核心技术的研究;赵亮;《中国优秀硕士学位论文全文数据库》;20120531;I136-368 * |
| 利用变换域信息快速实现H.264帧内预测编码的新算法;苏睿，刘贵忠，张彤宇，刘宝兰;《电子与信息学报》;20070131;第29卷(第1期);161-165 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103997646A (en) | 2014-08-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103997646B (en) | Fast intra-mode prediction mode selecting method in a kind of HD video coding | |
| CN109688414B (en) | VVC intra-frame coding unit candidate prediction mode reduction and block division early termination method | |
| KR100952340B1 (en) | Method and Apparatus for Determing Encoding Mode by Using Temporal and Spartial Complexity | |
| KR100739714B1 (en) | Method and apparatus for intra prediction mode decision | |
| US8144770B2 (en) | Apparatus and method for encoding moving picture | |
| CN101394565B (en) | Intra-frame prediction method | |
| US10390020B2 (en) | Video encoding methods and systems using adaptive color transform | |
| CN107087200B (en) | Skip coding mode advanced decision method for high-efficiency video coding standard | |
| KR20130004133A (en) | Method and apparatus for predicting chrominence component image uisng luminance component image | |
| CN106254870B (en) | Video encoding method, system and computer-readable recording medium using adaptive color conversion | |
| KR20130002284A (en) | Method and apparatus for video encoding with adjusting offsets based on pixel classification, method and apparatus for video decoding with adjusting offsets based on pixel classification | |
| US8189667B2 (en) | Moving picture encoding apparatus | |
| CN104052994A (en) | Method for HEVC intra-frame prediction mode quick decision based on grading self-adaption | |
| US9294764B2 (en) | Video encoder with intra-prediction candidate screening and methods for use therewith | |
| CN111837389A (en) | Block detection method and device suitable for multi-sign bit hiding | |
| KR20090075767A (en) | Method of encoding and decoding using the directional information of the prediction mode of the adjacent blocks for estimating the intra prediction mode, and apparatus thereof | |
| KR20130068479A (en) | Method of encoding intra mode by choosing most probable mode with high hit rate and apparatus for the same, and method of decoding and apparatus for the same | |
| CN103384326A (en) | Quick intra-frame prediction mode selection method for AVS-M video coding | |
| CN114339218A (en) | Image encoding method, image encoding device, electronic apparatus, and readable storage medium | |
| KR101902700B1 (en) | Apparatus for hevc coding and method for process of coding in intra prediction using the same | |
| EP3104606A1 (en) | Video encoding methods and systems using adaptive color transform | |
| CN110913232B (en) | Selection method and device of TU division mode and readable storage medium | |
| CN106878754A (en) | A kind of 3D video depths image method for choosing frame inner forecast mode | |
| CN113906754A (en) | Image encoding/decoding method and apparatus | |
| Li et al. | Content based fast intra coding for AVS2 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20191022 Address after: 610000 Chengdu Free Trade Experimental Zone, Sichuan Province, No. 300 Jiaozi Avenue, Chengdu High-tech Zone, No. 3 Building, 22 Floor 5 Patentee after: CHENGDU VISION-ZENITH TECHNOLOGY DEVELOPMENT CO., LTD. Address before: 100191 Haidian District, Xueyuan Road, No. 37, Patentee before: Beijing University of Aeronautics and Astronautics |