CN103533369B - A kind of fast schema selection method of gradable video encoding spatial enhancement layer - Google Patents
A kind of fast schema selection method of gradable video encoding spatial enhancement layer Download PDFInfo
- Publication number
- CN103533369B CN103533369B CN201310508345.3A CN201310508345A CN103533369B CN 103533369 B CN103533369 B CN 103533369B CN 201310508345 A CN201310508345 A CN 201310508345A CN 103533369 B CN103533369 B CN 103533369B
- Authority
- CN
- China
- Prior art keywords
- mode
- coding
- enhancement layer
- layer
- correlation
- 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.)
- Expired - Fee Related
Links
- 238000010187 selection method Methods 0.000 title claims abstract description 14
- 239000010410 layer Substances 0.000 claims abstract description 115
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 34
- 238000000034 method Methods 0.000 claims abstract description 33
- 239000011229 interlayer Substances 0.000 claims abstract description 29
- 238000013139 quantization Methods 0.000 claims abstract description 23
- 230000008569 process Effects 0.000 claims description 10
- 230000006835 compression Effects 0.000 claims description 8
- 238000007906 compression Methods 0.000 claims description 8
- 238000011160 research Methods 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 6
- 230000009191 jumping Effects 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 claims description 2
- 238000005457 optimization Methods 0.000 abstract description 3
- 230000009467 reduction Effects 0.000 description 6
- 239000013598 vector Substances 0.000 description 6
- 238000012360 testing method Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 238000012896 Statistical algorithm Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012913 prioritisation Methods 0.000 description 1
- 230000016776 visual perception Effects 0.000 description 1
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention discloses the fast schema selection method of a kind of gradable video encoding spatial enhancement layer, be mainly used for the novel spatial enhancement layer fast schema selection method of H.264 gradable video encoding (SVC) standard.The present invention proposes a kind of novel spatial enhancement layer fast schema selection method for H.264 gradable video encoding (SVC) standard.In order to improve code efficiency, on the basis of H.264/AVC, H.264/SVC add three kinds of inter-layer predication techniques.But, this considerably increases the complexity of SVC mode selection algorithm, have impact on the practicality of SVC standard.The algorithm that the present invention proposes takes full advantage of dependency, the dependency of interlayer quantization parameter and the dependency of adjacent macroblocks Coding cost between spatial enhancement layer macro block and corresponding Primary layer macro block, while keeping code efficiency, effectively reducing the computational complexity of mode selection algorithm, the optimization for SVC encoder provides theoretical reference value and actual application value.
Description
Technical Field
The invention relates to a rapid mode selection method of a scalable video coding spatial enhancement layer, which is mainly used for a novel rapid mode selection method of the spatial enhancement layer of an H.264 Scalable Video Coding (SVC) standard.
Background
Abbreviations and key term definitions:
scalable Video Coding
AVC Advanced video coding
QP Quantization Parameter
JSVM Joint Scalable Video Model Joint extensible Video Model
RDO Rate-Distortion Optimization
RDCost Rate-Distortion cost
GOP Group of Picture Group of pictures
PSNR Peak Signal-to-noise Ratio
BD-PSNR Bjontegaard delta PSNR
BD-BR Bjontegaard delta bitrate
ATR Average Time Reduction Average encoding Time Reduction
The Scalable Video Coding (SVC) standard is based on H.264/AVC (advanced Video Coding)[1]The video compression technology of (2) is a solution for coping with the heterogeneity of a transmission system and the diversity of terminals. The method has three forms of time gradable, space gradable and quality gradable, and can generate coding code streams with different frame rates, different code rates and different video qualities through one-time coding. At present, the coding efficiency of SVC is improved compared to the former standard, but at the same time, the coding complexity is also greatly increased. How to optimize an SVC encoder to increase the encoding speed according to the characteristics of SVC is a current research focus.
In SVC spatial scalable coding, the base layer coding follows two Intra (Intra) prediction modes (Intra4 × 4, Intra16 × 16) and seven Inter (Inter) prediction modes (16 × 16, 16 × 8, 8 × 16, 8 × 8, 8 × 4, 4 × 8, 4 × 4) that are already known in the h.264/AVC technology. For the enhancement layer, in order to fully utilize the correlation between spatial layers and reduce the inter-layer redundancy, under the original prediction mode of h.264/AVC, SVC adds three prediction modes: inter-layer intra prediction, inter-layer motion prediction, and inter-layer residual prediction. The use of multiple modes and the use of inter-layer prediction tools greatly improves coding efficiency and rate-distortion performance. However, the conventional mode selection method is to perform full search on all modes under all possible inter-layer prediction options to obtain an optimal mode, and the operation complexity is very high. With the expansion of the coding layer number, the operation complexity is multiplied. Therefore, fast algorithm for mode selection is an important content in the research of current SVC encoder.
Fast algorithms for spatially scalable enhancement layers are a research hotspot and have many research achievements. Document [2] selects an Intra/Inter type using search results of 8 × 8 and Intra4 × 4, and then screens candidate modes according to the macroblock mode of the base layer. Document [3] designs a lookup table between macroblock candidate mode, base layer macroblock mode, and Quantization Parameter (QP). Document [4] proposes predicting candidate modes from the base layer and spatial neighboring macroblocks and determining whether to exit the mode search early by comparing with a rate-distortion cost threshold. Document [5] predicts candidate modes using neighboring macroblock mode information at a corresponding position of the base layer.
However, the above algorithms have certain limitations. Document [2] needs to search all small block modes such as 8 × 8 forcibly, and the complexity is still high; in the document [3], the candidate modes are obtained by table lookup and cannot be adjusted according to the actual coding condition; document [5] requires multiple threshold comparisons, and the coding effect depends on the threshold setting.
[1]SEGALL C A,SULLIVAN G J.Spatial scalability within the H.264/AVCscalable video coding extensions[J].IEEE Transactions on Circuits and Systemsfor Video Technology，2007，17(9):1121－1135.
[2]LI H，LI Z G，WEN C.Fast mode decision algorithm for inter-framecoding in fully scalable video coding[J].IEEE Transactions on CircuitsandSystems for Video Technology，2006，16(7):889－895.
[3]LIN H C，PENG W H，HANG H M，et al.Layer-adaptive mode decision andmotion search for scalable video coding with combined coarse granularscalability(CGS)and temporal scalability[C]//Proceedings of InternationalConference on Image Processing(ICIP).San Antonio:IEEE，2007:II289－II292.
[4]REN J，KEHTARNAVAZ N.Fast adaptive termination mode selection forH.264scalable video coding[J].Journal of Real-Time Image Processing，2009，4(1):13－21.
[5]KIM S T，KONDA K R，PARK C S，et al.Fast mode decision algorithm forinter-layer coding in scalable video coding[J].IEEE Transactions on ConsumerElectronics，2009，55(3):1572－1580.
[6]CHANG K，QIN T F，ZHENG J L，et al.Fast mode decision for scalablevideo coding with spatial scalability[C]//WiCOM 2011:Proceedings of the 7thInternational Conference on Wireless Communications，Networking and MobileComputing.Piscataway:IEEE Press，2011:1－4.
Disclosure of Invention
The purpose of the invention is as follows: in view of the problems and deficiencies in the prior art, the present invention provides a method for fast mode selection of a scalable video coding spatial enhancement layer, which adaptively reduces the computational complexity of a spatial enhancement layer macroblock mode selection process by utilizing the correlation between a spatial enhancement layer macroblock and a corresponding base layer macroblock, the correlation of inter-layer quantization parameters, and the correlation of coding costs of adjacent macroblocks. By adopting the method of the invention, the H.264/SVC spatial hierarchical coding speed can be greatly improved in a large QP range, and meanwhile, the operation complexity of the mode selection algorithm is effectively reduced on the premise of keeping the coding efficiency.
The technical scheme is as follows: a method for selecting fast mode of scalable video coding spatial enhancement layer, which utilizes the correlation between spatial enhancement layer macro block and corresponding base layer macro block, the correlation of inter-layer quantization parameter and the correlation of adjacent macro block coding cost to reduce the calculation complexity of the spatial enhancement layer macro block mode selection process, the specific scheme is as follows:
1 accelerating mode selection using correlation of inter-layer modes
Much of the research on inter-mode decision algorithms in SVC spatial scalable coding is based on statistical algorithms that account for the best coding mode of macroblocks in the base and enhancement layers[6]From these statistics, it can be seen that there is some commonality that the macroblock prediction modes of the base layer and the enhancement layer have strong correlation.
When the macroblock in the base layer adopts the Intra coding mode, the corresponding candidate mode sets of the macroblock in the enhancement layer include three modes, i.e., Intra bl, Intra4 × 4, and Intra16 × 16. Document [2] shows that the Intra16 × 16 mode is rarely adopted by observing a large number of sequences, so the Intra mode candidate set can be reduced to the Intra bl and Intra4 × 4 modes, which saves time and ensures coding performance.
When the macroblock in the base layer adopts the Inter coding mode, if the base layer selects the SKIP or the Inter16 × 16, Inter16 × 8, Inter8 × 16, etc. large-size macroblock mode as the best coding mode, it indicates that the image is the background area or the area with not severe motion in this area, and the enhancement layer preferably selects the same mode or the similar mode for coding when the corresponding macroblock of the enhancement layer is coded. When the base layer uses Inter8 × 8 macroblocks as the best mode for encoding, it is indicated that this region is a region with more texture details or more intense motion, and in order to ensure the encoding quality, the Inter8 × 8 mode is also mostly used as the best mode for the corresponding macroblocks of the enhancement layer. In addition, the algorithm always keeps two modes, namely a DIRECT mode and a BLSkip mode, as candidate coding modes in general; except that the probability of the enhancement layer selecting the Inter8 × 8 mode is larger when the base layer selects the Inter8 × 8 mode, the probability of the enhancement layer selecting the Inter8 × 8 mode is small under other conditions, therefore, in the algorithm, the enhancement layer selects the Inter8 × 8 mode as the candidate mode only when the base layer selects the Inter8 × 8 mode. In addition, the DIRECT mode directly obtains coding information such as motion vectors from adjacent coded macro blocks by using time and space correlation, the BLSkip mode obtains coding information such as motion segmentation, motion vectors, reference frame indexes and the like from corresponding coded macro blocks of a base layer by using interlayer correlation, and the coding complexity can be effectively reduced without searching and matching optimal motion vectors, so that two modes of DIRECT and BLSkip are always reserved as candidate modes in the algorithm.
2 accelerating mode selection using correlation of inter-layer quantization parameters
The respective picture quality of the base layer and the enhancement layer will also have an impact on the distribution of their macroblock coding modes. Let QPBLAnd QPELRepresenting the quantization parameters of the base layer and the enhancement layer, respectively, and dQP represents the difference between the quantization parameters of the base layer and the enhancement layer:
dQP has an impact on the spatial enhancement layer macroblock coding mode. First, all macroblock coding modes are divided into three subsets, namely a Coarse class (Coarse), a Fine class (Fine) and a Skip class (Skip), different sequences are coded by adopting a 'full traversal' mode selection method of JSVM by using different dQPs, and the probability that the optimal modes of all macroblocks fall into different subsets is counted, so that as the dQP is increased from-10 to 10, the Skip class mode is more and more used, and the Coarse and Fine class modes are less and less, specifically:
1. when dQP is positive and the absolute value is large, the Skip class mode is dominant, and the Fine class tends to be 0;
2. when dQP is near 0, the probability of Coarse and Fine class modes is greater than that of Skip class;
3. when dQP is negative and the absolute value is large, more than half of the macroblocks adopt the Coarse mode.
Therefore, the order of RDO mode selection is adaptively adjusted by effectively utilizing the dQP information, that is, the mode selected as the mode with the highest probability of the final mode should be preferentially calculated for RDCost, which also plays a positive role in rapidly realizing coding.
3 accelerating mode selection by using correlation of adjacent macroblock coding cost
In the process of coding the image by adopting the RDO method, the RD cost of adjacent macro blocks has strong correlation[4]And therefore can be used for implementation of fast coding. If RD costs of left, upper and upper left macroblocks of the current coding macroblock are RDA、RDBAnd RDCThen a threshold function TH of the RD cost may be definedRDAs shown in the following formula:
in the formula, median (-) is a median operator, α has a value range of 0-1, and a proper value α can make a compromise between computational complexity and compression efficiencyValue THRDThen the current mode can be considered to be accurate enough, and therefore the RD cost calculation for the subsequent mode need not be solved. If a large TH is usedRDRDO cycles can jump out quickly to speed up mode selection; if a small TH is usedRDThe final mode selection result will be more accurate but the complexity will increase.
Based on the above conclusion, when the available mode is calculated in a traversal mode, if the result of the current mode meets the early termination condition, namely the rate-distortion cost is less than THRDIf all the available modes do not meet the condition of early termination, the mode corresponding to the minimum rate distortion cost is the final mode of the current macro block, a large number of experimental simulation results show that the value of α should be different for video sequences with different resolution sizes, when the spatial resolution of the spatial enhancement layer is greater than 4CIF, the value of α should be 0.35, and when the spatial resolution of the spatial enhancement layer is less than CIF, the value of α should be 0.25.
The core method of the fast mode selection method is to utilize the determinacy of the quantization parameter to the mode distribution to reduce the range of the available modes and utilize the related characteristic of the rate distortion cost of the adjacent macro blocks to terminate the traversal of the modes in advance, thereby achieving the purpose of improving the coding speed on the premise of keeping the coding compression efficiency.
Has the advantages that: in order to improve the coding efficiency, three inter-layer prediction technologies are added to H.264/AVC in H.264/SVC, so that the number of modes available when encoding the H.264/SVC spatial scalable enhancement layer macro block is about 2 times that of H.264/AVC encoding. The more coding modes that can be selected means the higher the computational complexity of mode selection based on rate-distortion optimization, which greatly increases the complexity of the SVC mode selection algorithm and affects the practicality of the SVC standard. And the mode selection of the coding of the macroblock of the spatial enhancement layer of the H.264/SVC official reference model JSVM adopts a 'full traversal' mode, namely, all available modes are coded once, and finally, the optimal mode is selected as the final coding mode by obtaining the minimum RD cost. The "full traversal" method can obtain the optimal coding compression efficiency, but the computation complexity is very high.
The invention provides a novel H.264/SVC space enhancement layer fast mode selection algorithm aiming at the defect of a 'full traversal' mode in JSVM. The adaptive reduction of the computational complexity of the spatial enhancement layer macroblock mode selection process is achieved by utilizing the correlation between the spatial enhancement layer macroblock and the corresponding base layer macroblock, the correlation of the inter-layer quantization parameter, and the correlation of the coding cost of the neighboring macroblocks. By adopting the method of the invention, the H.264/SVC spatial hierarchical coding speed can be greatly improved in a large QP range, and meanwhile, the operation complexity of the mode selection algorithm is effectively reduced on the premise of keeping the coding efficiency.
Drawings
FIG. 1 is a flow chart of a method based on inter-layer pattern correlation;
fig. 2 is a flowchart of a fast mode selection method according to the present invention.
Detailed Description
The present invention is further illustrated by the following examples, which are intended to be purely exemplary and are not intended to limit the scope of the invention, as various equivalent modifications of the invention will occur to those skilled in the art upon reading the present disclosure and fall within the scope of the appended claims.
The method for selecting the fast mode of the scalable video coding spatial enhancement layer reduces the computational complexity of the spatial enhancement layer macro block mode selection process by utilizing the correlation between the spatial enhancement layer macro block and the corresponding base layer macro block, the correlation of the interlayer quantization parameter and the correlation of the coding cost of the adjacent macro blocks, and the specific scheme is as follows:
1 accelerating mode selection using correlation of inter-layer modes
Much of the research on inter-mode decision algorithms in SVC spatial scalable coding is based on statistical algorithms that account for the best coding mode of macroblocks in the base and enhancement layers[6]From these statistics, it can be seen that there is some commonality that the macroblock prediction modes of the base layer and the enhancement layer have strong correlation.
When the macroblock in the base layer adopts the Intra coding mode, the corresponding candidate mode sets of the macroblock in the enhancement layer include three modes, i.e., Intra bl, Intra4 × 4, and Intra16 × 16. Document [2] shows that the Intra16 × 16 mode is rarely adopted by observing a large number of sequences, so the Intra mode candidate set can be reduced to the Intra bl and Intra4 × 4 modes, which saves time and ensures coding performance.
When the macroblock in the base layer adopts the Inter coding mode, if the base layer selects the SKIP or the Inter16 × 16, Inter16 × 8, Inter8 × 16, etc. large-size macroblock mode as the best coding mode, it indicates that the image is the background area or the area with not severe motion in this area, and the enhancement layer preferably selects the same mode or the similar mode for coding when the corresponding macroblock of the enhancement layer is coded. When the base layer uses Inter8 × 8 macroblocks as the best mode for encoding, it is indicated that this region is a region with more texture details or more intense motion, and in order to ensure the encoding quality, the Inter8 × 8 mode is also mostly used as the best mode for the corresponding macroblocks of the enhancement layer. In addition, the algorithm always keeps two modes, namely a DIRECT mode and a BLSkip mode, as candidate coding modes in general; except that the probability of the enhancement layer selecting the Inter8 × 8 mode is larger when the base layer selects the Inter8 × 8 mode, the probability of the enhancement layer selecting the Inter8 × 8 mode is small under other conditions, therefore, in the algorithm, the enhancement layer selects the Inter8 × 8 mode as the candidate mode only when the base layer selects the Inter8 × 8 mode. In addition, the DIRECT mode directly obtains coding information such as motion vectors from adjacent coded macro blocks by using time and space correlation, the BLSkip mode obtains coding information such as motion segmentation, motion vectors, reference frame indexes and the like from corresponding coded macro blocks of a base layer by using interlayer correlation, and the coding complexity can be effectively reduced without searching and matching optimal motion vectors, so that two modes of DIRECT and BLSkip are always reserved as candidate modes in the algorithm.
The flow of the algorithm based on inter-layer pattern correlation is shown in fig. 1.
2 accelerating mode selection using correlation of inter-layer quantization parameters
The respective picture quality of the base layer and the enhancement layer will also have an impact on the distribution of their macroblock coding modes. Let QPBLAnd QPELRepresenting the quantization parameters of the base layer and the enhancement layer, respectively, and dQP represents the difference between the quantization parameters of the base layer and the enhancement layer:
dQP has an impact on the spatial enhancement layer macroblock coding mode. All macroblock coding modes are first divided into three subsets, namely Coarse (Coarse), Fine (Fine) and Skip (Skip), as shown in table 1:
table 1 classification of enhancement layer macroblock coding modes
Using a "full traversal" mode selection method using JSVM for different dQP to encode different sequences and counting the probability that the optimal mode for all macroblocks falls within different subsets, it can be seen that as dQP grows from-10 to 10, Skip class mode is used more and more, while Coarse and Fine class modes are less and less, specifically:
1. when dQP is positive and the absolute value is large, the Skip class mode is dominant, and the Fine class tends to be 0;
2. when dQP is near 0, the probability of Coarse and Fine class modes is greater than that of Skip class;
3. when dQP is negative and the absolute value is large, more than half of the macroblocks adopt the Coarse mode.
Therefore, the order of RDO mode selection is adaptively adjusted by effectively utilizing the dQP information, that is, the mode selected as the mode with the highest probability of the final mode should be preferentially calculated for RDCost, which also plays a positive role in rapidly realizing coding.
Based on the above conclusion, the coding mode should be calculated according to different dQP in different orders, i.e. the mode with high probability of being the final optimal mode is calculated first. The strategy can achieve the purpose of rapid convergence of mode selection by matching with the early termination condition. Table 2 shows a mode priority ordering method based on dQP values:
TABLE 2 prioritization of modes under different dQP (1 is highest priority)
3 accelerating mode selection by using correlation of adjacent macroblock coding cost
In the process of coding the image by adopting the RDO method, the RD cost of adjacent macro blocks has strong correlation[4]And therefore can be used for implementation of fast coding. If let RD cost of left, upper and upper left macroblocks of current coding macroblock be dividedIs otherwise RDA、RDBAnd RDCThen a threshold function TH of the RD cost may be definedRDAs shown in the following formula:
in the formula, median (-) is a median operator, α has a value range of 0-1, and a proper α value can make a compromise between computational complexity and compression efficiencyRDThen the current mode can be considered to be accurate enough, and therefore the RD cost calculation for the subsequent mode need not be solved. If a large TH is usedRDRDO cycles can jump out quickly to speed up mode selection; if a small TH is usedRDThe final mode selection result will be more accurate but the complexity will increase.
Based on the above conclusion, when the available mode is calculated in a traversal mode, if the result of the current mode meets the early termination condition, namely the rate-distortion cost is less than THRDAnd then jumping out of the traversal process, wherein the current mode is the final mode of the current macro block. And if all the available modes do not meet the early termination condition, after the traversal is finished, the mode corresponding to the minimum rate distortion cost is the final mode of the current macro block. A large number of experimental simulation results show that videos with different resolution ratiosIn the algorithm, α should take a value of 0.35 when the spatial resolution of the spatial enhancement layer is greater than 4CIF, and α should take a value of 0.25 when the spatial resolution of the spatial enhancement layer is less than CIF.
By combining the results of the above three research and analysis, a fast mode selection algorithm for coding an h.264/SVC spatial enhancement layer macroblock by comprehensively utilizing inter-layer mode correlation, inter-layer quantization parameter correlation, and neighboring macroblock coding cost correlation can be proposed, and the flow is shown in fig. 2. Firstly, according to different dQPs, a mode subset with high optimal mode probability is selected for calculation preferentially. And finally, determining the range of the available modes by utilizing the determinacy of the base layer quantization parameter to the mode distribution, determining the alternative modes, and calculating the rate-distortion cost RDcost of each alternative mode according to the range. If the current mode meets the early termination condition, determining that the current mode is the optimal mode, and jumping out of the calculation process; and if the current mode does not meet the early termination condition, continuously calculating the RDcost of the remaining alternative modes, and judging whether the current mode meets the early termination condition or not. And if all the alternative modes do not meet the early termination condition, selecting the mode with the minimum rate distortion cost as the optimal mode. The core method of the rapid mode selection algorithm is to utilize the determinacy of quantization parameters on mode distribution to reduce the range of available modes and utilize the correlation characteristic of rate distortion cost of adjacent macro blocks to terminate the traversal of the modes in advance, thereby achieving the purpose of improving the coding speed on the premise of keeping the coding compression efficiency.
The method is realized based on an H.264/SVC official test model JSVM, and a series of video sequences with different resolution and image complexity are selected in an experiment to carry out coding test, wherein the video sequences comprise 4 sequences including Foreman, Mobile, Football and City. All experimental code streams are composed of a base layer and an enhancement layer, and the width and the height of the enhancement layer are 2 times of those of the base layer, namely the spatial hierarchy of the base two. In all experiments, the base layer format is qcif (quart common Intermediate format), the enhancement layer format is cif (common Intermediate format), the frame rate is set to 15 frames per second, the Group of pictures (GOP) size is set to 8, and each sequence is encoded over 100 frames to eliminate uncertainty.
Parameters for evaluating the coding efficiency and coding complexity of the algorithm include: BD-PSNR (Bjontegaarddelta PSNR), BD-BR (Bjontegaard delta bitrate), and average coding time reduction (ATR).
The BD-PSNR and the BD-BR can represent the average difference of the PSNR and the bit rate between the JSVM algorithm and the RD curve of the algorithm by utilizing a curve fitting principle, and can be used for reflecting the global coding efficiency gain of the algorithm. ATR is used to illustrate the degree of complexity reduction of the new algorithm relative to the "full traversal" mode selection algorithm of JSVM, which is expressed as follows:
wherein, TimeJSVMAnd TimenewRespectively representing the JSVM algorithm and the coding time of the algorithm.
The results of the experiment are shown in table 3. Where negative values of BD-PSNR indicate a decrease in quality, positive values of BD-BR indicate an increase in bit rate, and positive values of ATR indicate a decrease in coding time.
TABLE 3 comparison of the algorithm of the present invention with that of JSVM
As can be seen from the data in Table 3, the method proposed by the present invention is effective for all 4 test sequences, i.e., Foreman, Mobile, Football, City. Comparing simulation results of the two methods, the complete method has the advantages that compared with the JSVM algorithm, the coding bit rate is slightly improved, but the change is less than 2%, the coding quality is slightly reduced, but the coding quality is less than 0.2dB, and the subjective visual perception is basically unchanged. In addition, for the 4 test sequences, the reduction of the coding time can reach more than 50 percent, thereby greatly saving the coding time. Moreover, the coding performance of the present algorithm is relatively stable for different QP values.
Claims (3)
1. A method for fast mode selection for a spatial enhancement layer for scalable video coding, comprising: the method comprises three parts of utilizing the correlation of an interlayer mode to accelerate mode selection, utilizing the correlation of an interlayer quantization parameter to accelerate mode selection and utilizing the correlation of adjacent macroblock coding cost to accelerate mode selection;
wherein,
accelerating mode selection using inter-layer mode correlation
When the macro block in the basic layer adopts an Intra-frame coding mode, reducing an Intra-frame mode candidate set into an IntraBL mode and an Intra4 × 4 mode, when the macro block in the basic layer adopts the inter-frame coding mode, if the basic layer selects a large-size macro block mode as an optimal coding mode, indicating that an image in the area is a background area or an area with non-violent motion, and preferentially selecting the same mode or a similar mode for coding when the corresponding macro block of an enhancement layer is coded; when the base layer adopts an Inter8 × 8 macroblock as the best mode for encoding, it indicates that the region is a region with more texture details or more intense motion, and the corresponding macroblock of the enhancement layer also adopts an Inter8 × 8 macroblock as the best mode; the algorithm always keeps a DIRECT mode and a BLSkip mode as candidate coding modes; only when the base layer selects an Inter8 × 8 mode, the enhancement layer selects an Inter8 × 8 mode as a candidate mode;
accelerating mode selection using correlation of inter-layer quantization parameters
Let QPBLAnd QPELRepresenting the quantization parameters of the base layer and the enhancement layer, respectively, and dQP represents the difference between the quantization parameters of the base layer and the enhancement layer:
the dQP has influence on the macroblock coding mode of the spatial enhancement layer, firstly, all the macroblock coding modes are divided into three subsets, namely a coarse class, a fine class and a skip class, different sequences are coded by adopting different dQPs and using a JSVM full traversal mode selection method, and the probability that the optimal modes of all the macroblocks fall into different subsets is counted; the coding mode should be calculated according to different dQPs in different orders;
correlation-based accelerated mode selection using neighboring macroblock coding costs
In the process of encoding an image by using the RDO method, if the RD cost of the left, upper and upper left macroblocks of the current encoding macroblock is madeAre each RDA、RDBAnd RDCThen a threshold function TH of the RD cost may be definedRDAs shown in the following formula:
in the formula, median (-) is a median operator, α has a value range of 0-1, and proper α value can make a compromise between computational complexity and compression efficiency, if the RD cost of the current macroblock in a certain mode is less than a threshold THRDThen, the current mode can be considered to be accurate enough, and therefore, the RD cost calculation of the subsequent mode is not required to be obtained; if a large TH is usedRDRDO cycles can jump out quickly to speed up mode selection; if a small TH is usedRDThe final mode selection result will be more accurate but the complexity will increase; when the available mode is calculated in a traversal mode, if the result of the current mode meets the early termination condition, namely the rate distortion cost is less than THRDIf yes, jumping out of the traversal process, wherein the current mode is the final mode of the current macro block; and if all the available modes do not meet the early termination condition, after the traversal is finished, the mode corresponding to the minimum rate distortion cost is the final mode of the current macro block.
2. The method of fast mode selection for a spatial enhancement layer for scalable video coding as claimed in claim 1 wherein: when the spatial resolution of the spatial enhancement layer is greater than 4CIF, alpha should be 0.35; when the spatial resolution of the spatial enhancement layer is smaller than CIF, α should be 0.25.
3. The method of fast mode selection for a spatial enhancement layer for scalable video coding as claimed in claim 1 wherein: according to the research and analysis results of the three parts, the fast mode selection method of H.264/SVC spatial enhancement layer macro block coding comprehensively utilizes the inter-layer mode correlation, the inter-layer quantization parameter correlation and the adjacent macro block coding cost correlation, utilizes the determinacy of the quantization parameter to the mode distribution to reduce the range of the available mode, and utilizes the related characteristic of the adjacent macro block rate distortion cost to terminate the traversal of the mode in advance, thereby achieving the purpose of improving the coding speed on the premise of keeping the coding compression efficiency.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310508345.3A CN103533369B (en) | 2013-10-24 | 2013-10-24 | A kind of fast schema selection method of gradable video encoding spatial enhancement layer |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310508345.3A CN103533369B (en) | 2013-10-24 | 2013-10-24 | A kind of fast schema selection method of gradable video encoding spatial enhancement layer |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103533369A CN103533369A (en) | 2014-01-22 |
| CN103533369B true CN103533369B (en) | 2016-10-26 |
Family
ID=49934982
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310508345.3A Expired - Fee Related CN103533369B (en) | 2013-10-24 | 2013-10-24 | A kind of fast schema selection method of gradable video encoding spatial enhancement layer |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103533369B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104768008B (en) * | 2015-03-26 | 2018-01-30 | 哈尔滨工业大学 | A kind of fast encoding method for scalable video |
| CN104954785A (en) * | 2015-06-16 | 2015-09-30 | 哈尔滨工业大学 | Layered mode decision method used for scalable video coding |
| WO2022141282A1 (en) * | 2020-12-30 | 2022-07-07 | 深圳市大疆创新科技有限公司 | Video processing method and encoding device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101478671A (en) * | 2008-01-02 | 2009-07-08 | 中兴通讯股份有限公司 | Video encoding apparatus applied on video monitoring and video encoding method thereof |
| CN101860759A (en) * | 2009-04-07 | 2010-10-13 | 华为技术有限公司 | Encoding method and encoding device |
| CN102088605A (en) * | 2011-02-23 | 2011-06-08 | 浙江大学 | Rapid interlayer prediction selection method for scalable video coding |
| CN102740078A (en) * | 2012-07-12 | 2012-10-17 | 北方工业大学 | Adaptive spatial scalable coding based on HEVC (high efficiency video coding) standard |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8654835B2 (en) * | 2008-09-16 | 2014-02-18 | Dolby Laboratories Licensing Corporation | Adaptive video encoder control |
-
2013
- 2013-10-24 CN CN201310508345.3A patent/CN103533369B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101478671A (en) * | 2008-01-02 | 2009-07-08 | 中兴通讯股份有限公司 | Video encoding apparatus applied on video monitoring and video encoding method thereof |
| CN101860759A (en) * | 2009-04-07 | 2010-10-13 | 华为技术有限公司 | Encoding method and encoding device |
| CN102088605A (en) * | 2011-02-23 | 2011-06-08 | 浙江大学 | Rapid interlayer prediction selection method for scalable video coding |
| CN102740078A (en) * | 2012-07-12 | 2012-10-17 | 北方工业大学 | Adaptive spatial scalable coding based on HEVC (high efficiency video coding) standard |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103533369A (en) | 2014-01-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109688414B (en) | VVC intra-frame coding unit candidate prediction mode reduction and block division early termination method | |
| CN101964906B (en) | Rapid intra-frame prediction method and device based on texture characteristics | |
| CN103248893B (en) | From H.264/AVC standard to code-transferring method and transcoder thereof the fast frame of HEVC standard | |
| CN104539962B (en) | It is a kind of merge visually-perceptible feature can scalable video coding method | |
| KR101433170B1 (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 | |
| CN108989802B (en) | HEVC video stream quality estimation method and system by utilizing inter-frame relation | |
| JP2006519565A (en) | Video encoding | |
| TW201004357A (en) | Rate-distortion quantization for context-adaptive variable length coding (CAVLC) | |
| CN101247525B (en) | Method for improving image intraframe coding velocity | |
| AU2006223416A1 (en) | Content adaptive multimedia processing | |
| WO2010078759A1 (en) | Method for image temporal and spatial resolution processing based on code rate control | |
| JP2008541570A (en) | Rate control for multi-layer video design | |
| CN106210721B (en) | A kind of quick code check code-transferring methods of HEVC | |
| CN101494792A (en) | H.264/AVC frame inner prediction method based on edge characteristics | |
| CN101835042A (en) | Wyner-Ziv video coding system controlled on the basis of non feedback speed rate and method | |
| CN110996102A (en) | Video coding method and device for inhibiting intra-frame block respiration effect in P/B frame | |
| CN100579228C (en) | Method for rapidly predicting frame space of aerophotographic traffic video | |
| CN105681797A (en) | Prediction residual based DVC-HEVC (Distributed Video Coding-High Efficiency Video Coding) video transcoding method | |
| CN105187826A (en) | Rapid intra-frame mode decision method specific to high efficiency video coding standard | |
| CN103533369B (en) | A kind of fast schema selection method of gradable video encoding spatial enhancement layer | |
| CN108989818B (en) | Image coding parameter adjusting method and device | |
| CN100586186C (en) | Quick inter-frame forecast mode selection method | |
| CN101557519B (en) | Multi-view video coding method | |
| US20130128954A1 (en) | Encoding method and apparatus | |
| CN100586185C (en) | Mode selection method for transcoding 264 video to reduce resolving capability |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20161026 Termination date: 20191024 |