CN104333754B - Based on the SHVC enhancement-layer video coding methods that predictive mode is quickly selected - Google Patents
Based on the SHVC enhancement-layer video coding methods that predictive mode is quickly selected Download PDFInfo
- Publication number
- CN104333754B CN104333754B CN201410609900.6A CN201410609900A CN104333754B CN 104333754 B CN104333754 B CN 104333754B CN 201410609900 A CN201410609900 A CN 201410609900A CN 104333754 B CN104333754 B CN 104333754B
- Authority
- CN
- China
- Prior art keywords
- layer
- video
- candidate modes
- enhancement
- encoded
- 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
- 238000000034 method Methods 0.000 title claims abstract description 63
- 239000010410 layer Substances 0.000 claims description 217
- 238000013139 quantization Methods 0.000 claims description 14
- 238000005070 sampling Methods 0.000 claims description 8
- 239000011229 interlayer Substances 0.000 claims description 7
- 230000002708 enhancing effect Effects 0.000 claims description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 241000219357 Cactaceae Species 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000005728 strengthening Methods 0.000 description 1
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention discloses a kind of SHVC enhancement-layer video coding methods quickly selected based on predictive mode, enhancement layer coding speed is low in mainly solving the problems, such as scalable efficient video coding standard.Implementation step is：(1) determine the correlation of Primary layer and enhancement layer and obtain predictive mode probability statistics table by probability statistics；(2) the rough candidate modes of enhancement layer are selected according to probability statistics table；(3) n frames before video sequence are encoded respectively with rough candidate modes and Yin methods, and obtains distortion performance estimator BD PSNR；(4) rough candidate modes are adjusted according to distortion performance estimator BD PSNR and obtains final candidate modes；(5) video sequence is encoded with final candidate modes.The present invention significantly reduces encoder complexity on the premise of video distortion performance is ensured, reduces the scramble time, can be used for real-time video application.
Description
Technical field
The invention belongs to field of video processing, more particularly to a kind of predictive mode fast selecting method can be used for video pressure
Contracting.
Background technology
In January, 2013, by the big world Video coding groups of dynamic image expert group MPEG and Video Coding Experts Group VCEG two
Vertical Video coding joint working group JCT-VC exploitations are made into have formulated instead of International video coding new standard H.264/AVC
HEVC.But coding standard lacks flexibility, when resolution ratio, the frame per second of different terminals for video etc. has different requirements,
Just the video flowing being input into must repeatedly be encoded.In order to solve this problem, in the base of efficient video coding HEVC standard
Scalable video technology is expanded on plinth.
The vision signal of input can be disposably encoded into different code checks, space point by scalable efficient video coding SHVC
The code stream of resolution and video quality, to adapt to the requirement of different bandwidth and different clients to video, compares network requirement
It is low, share wires are used, moreover it is possible to realize the real-time, interactive of video conference, using more universal in video conference.
There is being used in combination for the retractilities such as time domain, spatial domain, quality and bit-depth and various retractilities in SHVC.Its
In, it refers to the video flowing for producing various spatial resolutions by first encoding on the basis of source code flow that spatial domain is scalable, and
And the height of the resolution ratio Primary layer of enhancement layer, but the picture material of each layer is identical, simply spatial resolution is higher, depending on
Frequency becomes apparent from exquisiteness.Original input video is compiled by down-sampling using efficient video coding HEVC or advanced videos
Code AVC is encoded into being Primary layer bit stream.The inter-layer reference image that the reconstructing video of Primary layer is obtained by up-sampling
Can be used for inter-layer prediction, enhancement layer coding quality can be improved using inter-layer prediction encoding enhancement layer image.
Because enhancement layer with the addition of inter-layer prediction mode ILR in SHVC, and all code tree units of enhancement layer will be carried out
Four codings of depth, each coding depth has various predictive modes, including SKIP, Inter_2N × 2N, Inter_2N × N,
Inter_N×2N、Inter_2N×nU、Inter_2N×nD、Inter_nL×2N、Inter_nR×2N、Intra_2N×2N、
Intra_N × N and ILR, it is necessary to a large amount of rate distortion costs are calculated, by can just obtain optimum prediction mode after rate-distortion optimization,
Therefore whole process computation complexity is very high, reduces code rate, limits the practical application of standard, it is necessary to further speed up
Predictive mode selection course.
So far, the predictive mode fast selecting method being had pointed out during SHVC spatial domains are scalable mainly Peng Yin,
Taoran Lu and Tao Chen et al. is proposed in January, 2013 in the 12nd International video meeting that JCT-VC is organized
Entitled " Inter-layer reference picture placement " JCTVC-L0174 files in the enhancing mentioned
Layer model fast selecting method, Yin methods are called in the present invention.Particular content is as follows：
(1) cycle tests Primary layer be I frames, enhancement layer be P frames under conditions of, have two kinds of systems of selection.The first side
Method is：Intra patterns and ILR patterns are only selected, other all inter-frame modes are not selected, and second method is：Only selection makes
With Inter_2N × 2N patterns of merge, intra patterns and ILR patterns；
(2) it is that I frames, enhancement layer are under conditions of B frames or Primary layer are non-I frames, relative to base in cycle tests Primary layer
For this layer prediction process, enhancement layer prediction has two changes：One is in Inter_2N × 2N patterns, to be arranged by estimation
Except the selection of inter-layer reference image, two is that ILR patterns are done after intra patterns.
In above-mentioned Yin methods, for strengthening under low time delay B, low time delay P in Fig. 1 and Stochastic accessing these three configuration conditions
The special frames of layer are the P frames of black, are encoded using Yin methods, and other frames of enhancement layer are compiled using scalable efficient video
Code SHVC methods are encoded, although this method can reduce the scramble time on the premise of distortion performance is ensured.But, should
Method still has multitude of video frame not using fast method, and cataloged procedure still needs to consume the plenty of time.
The content of the invention
Deficiency it is an object of the invention to be directed to above-mentioned prior art, proposes a kind of quickly to be selected based on predictive mode
The method of SHVC enhancement-layer videos coding, on the premise of distortion performance is ensured, the coding for reducing all picture format configurations is complicated
Degree and scramble time.
The technical scheme of the object of the invention is realized, is comprised the following steps：
1. a kind of SHVC enhancement-layer video coding methods quickly selected based on predictive mode, are comprised the following steps：
2. the corresponding relation of Primary layer coding unit and enhancement layer coding unit is determined：
1.1) in the spatial domain of scalable efficient video coding SHVC is scalable, the quantization step of Primary layer and enhancement layer is given
Long value, is input into a video sequence and carries out down-sampling, obtains two groups of different videos of resolution ratio, wherein, resolution ratio is small to be regarded
Frequency is Primary layer, and the big video of resolution ratio is enhancement layer；
1.2) the big coding unit of a 2N × 2N of Primary layer is divided into 4 lower Item units of N × N, this 4 N
The coding unit of × N corresponds to 4 coding units of adjacent 2N × 2N of enhancement layer, N=4,8,16,32 respectively；
(2) count a video sequence before n frames Primary layer optimum prediction mode and enhancement layer optimum prediction mode it is general
Rate relation, 40 ﹤ n ﹤ 60：
2.1) it is by size 4 × 4 subregions by N × N coding units encoded in Primary layer, each 4 × 4 region has
A kind of predictive mode, counts the most predictive mode of access times in these 4 × 4 regions, and using the pattern as Primary layer N ×
The optimum prediction mode of N coding units；
2.2) optimum prediction mode of N × N coding units of Primary layer and corresponding is recorded before a video sequence in n frames
The optimum prediction mode of enhancement layer 2N × 2N coding units, obtains Primary layer optimum prediction mode and enhancement layer optimum prediction mode
Probabilistic relation；
(3) repeat step (1) and step (2), obtain the Primary layer optimum prediction mode of multiple video sequences and increase respectively
The probabilistic relation of strong layer optimum prediction mode, and probabilistic relation to all video sequences does averagely, obtains predictive mode probability
Statistical form；
(4) in predictive mode probability statistics table, 11 kinds of enhancement layer prediction moulds of every kind of Primary layer optimum prediction mode correspondence
Formula, for every kind of Primary layer optimum prediction mode, by its corresponding 11 kinds of enhancement layer predictive mode according to probability from big to small
Order is arranged, if probability is more than 5% predictive mode and number is no more than 7 kinds, directly as rough candidate prediction
Pattern, otherwise, select probability preceding 7 kinds of candidate patterns high are used as rough candidate modes；
(5) the preceding n frames of video sequence are encoded with rough candidate modes, record code check and brightness peak are believed
Make an uproar ratio；
(6) the preceding n frames of video are encoded with Yin methods, records code check and brightness peak signal to noise ratio；
(7) the quantization step value of Primary layer and enhancement layer, repeat step (1) to step (6) are changed；
(8) according to step (5) to code check and brightness peak signal to noise ratio in step (7), distortion performance estimator is obtained
BD-PSNR；
(9) result according to step (8) judges whether rough candidate modes are final candidate modes：If
Meet -0.055dB ﹤ BD-PSNR ﹤ -0.045dB, then rough candidate modes are exactly final candidate modes, no
Then, rough candidate modes are adjusted, make its BD-PSNR meet require, and by adjustment after rough candidate prediction
Pattern is used as final candidate modes；
(10) according to final candidate modes, video sequence is encoded：
10.1) it is input into a video sequence and carries out down-sampling, obtains Primary layer and enhancement-layer video；
10.2) the first frame of Primary layer and enhancement layer is encoded using Yin methods；
10.3) video to Primary layer and enhancement layer since the second frame is encoded with different methods respectively：If base
This layer is non-I frames, then basic layer video is encoded using efficient video coding HEVC methods, records all coding units
Optimum prediction mode, and the final candidate modes of enhancement layer are found according to Primary layer optimum prediction mode, enhancement layer is regarded
Frequency is encoded；If Primary layer is I frames, step 10.4 is performed)；
10.4) configuration condition of video sequence is judged：If the configuration condition of video sequence is low time delay, using efficient
Video coding HEVC methods are encoded to basic layer video, record the optimum prediction mode of all coding units of Primary layer, root
The final candidate modes of enhancement layer are found according to Primary layer optimum prediction mode, enhancement-layer video is encoded；If depending on
The configuration condition of frequency sequence is Stochastic accessing, then basic layer video is encoded using efficient video coding HEVC methods, profit
Enhancement-layer video is encoded with Yin methods.
It is of the invention to be had the following advantages that compared with existing Yin methods：
(a) present invention using identical with enhancement-layer video content this feature of the scalable middle Primary layer in SHVC spatial domains, with general
Primary layer optimum prediction mode and the probabilistic relation of enhancement layer optimum prediction mode that the method for rate statistics is obtained, can determine that out most
Whole candidate modes, reduce the predictive mode number of enhancement layer, on the premise of distortion performance is ensured, reduce volume
Code complexity, improves code rate；
B () present invention has done predictive mode probability statistics due to the various picture formats to video sequence, and obtained corresponding
Final candidate modes, therefore, it is possible to be widely used in low time delay B, low time delay P and the configuration of three kinds of video sequences of Stochastic accessing
Condition.
Brief description of the drawings
Fig. 1 is the video frame type of the existing Primary layer under three kinds of video sequence configuration conditions and enhancement layer；
Fig. 2 is of the invention to realize flow chart.
Specific embodiment
Below in conjunction with accompanying drawing, embodiments of the present invention is described in detail.The present embodiment is with technical solution of the present invention
Premise is implemented, and gives detailed implementation method and specific operation process, but protection scope of the present invention be not limited to it is following
Embodiment.
The present invention is carried out on SHM6.1, and the configuration condition of video sequence has low time delay B, low time delay P and connects at random
The quantization step QP entered under three kinds, each case is as shown in table 1：
Quantization step under 1 three kinds of video sequence configuration conditions of table
| Video sequence configuration condition | Primary layer quantization step QP0 | Enhancement layer quantization step-length QP1 |
| Low time delay B | 20~25,26~29,30~33,34~40 | QP0+0~QP0+1, QP0+2~QP0+3 |
| Low time delay P | 20~25,26~29,30~33,34~40 | QP0+0~QP0+1, QP0+2~QP0+3 |
| Stochastic accessing | 20~25,26~30,31~34,35~40 | QP0+0~QP0+1, QP0+2~QP0+3 |
Reference picture 2, it is of the invention to realize that step is as follows：
Step one：Determine the corresponding relation of Primary layer coding unit and enhancement layer coding unit.
1.1) video test sequence information is provided, it is as shown in the table：
The specifying information of the video test sequence of table 2
1.2) video sequence configuration condition is low time delay B, quantization step QP0=(20~25), QP1=(QP0+0~QP0+
1) when, Traffic video sequences in input table 2 simultaneously carry out down-sampling, obtain resolution ratio be respectively 1280 × 800 and 2560 ×
1600 two groups of videos, wherein, resolution ratio is Primary layer for 1280 × 800 video, and resolution ratio is 2560 × 1600 video
It is enhancement layer；
1.3) the big coding unit of a 2N × 2N of Primary layer is divided into 4 lower Item units of N × N, this 4 N
The coding unit of × N corresponds to 4 coding units of adjacent 2N × 2N, N=4,8,16,32 in enhancement layer respectively.
Step 2：The Primary layer optimum prediction mode of 50 frames and enhancement layer optimum prediction before statistics Traffic video sequences
The probabilistic relation of pattern.
2.1) it is by size 4 × 4 subregions by N × N coding units encoded in Primary layer, each 4 × 4 region has
A kind of predictive mode, counts the most predictive mode of access times in these 4 × 4 regions, and using the pattern as Primary layer N ×
The optimum prediction mode of N coding units；
2.2) optimum prediction mode of N × N coding units and right in the Primary layer of 50 frames before record Traffic video sequences
The optimum prediction mode of the enhancement layer 2N × 2N coding units answered, obtains Primary layer optimum prediction mode and enhancement layer optimum prediction
The probabilistic relation of pattern.
Step 3：Repeat step one and step 2, the Primary layer for respectively obtaining other six video sequences in table 2 are most preferably pre-
The probabilistic relation of survey pattern and enhancement layer optimum prediction mode, i.e. PeopleOnStreet, Kimono, ParkScene,
Cactus, BasketballDrive, BQTerrace, and probabilistic relation to seven video sequences does averagely, obtains predicting mould
Formula probability statistics, such as table 3：
Under the low time delay B configuration conditions of table 3, Primary layer optimum prediction during QP0=(20~25) QP1=(QP0+0~QP0+1)
Probabilistic relation (%) between pattern and enhancement layer optimum prediction mode
In table 3,0~pattern of pattern 10 represent respectively SKIP, Inter_2N × 2N, Inter_2N × N, Inter_N × 2N,
Inter_2N × nU, Inter_2N × nD, Inter_nL × 2N, Inter_nR × 2N, Intra_2N × 2N, Intra_N × N and
ILR this 11 kinds of predictive modes.
Step 4：In the predictive mode probability statistics of table 3, for every kind of Primary layer optimum prediction mode, corresponded to
11 kinds of enhancement layer predictive modes arranged according to probability order from big to small, if predictive mode and number of the probability more than 5%
No more than 7 kinds, then directly as rough candidate modes, otherwise, select probability preceding 7 kinds of candidate pattern conducts high
Rough candidate modes.
Step 5：50 frames before seven video sequences in table 2 are encoded respectively with rough candidate modes.
5.1) basic layer video is encoded with efficient video coding HEVC, records the optimum prediction of all coding units
Pattern；
5.2) according to the size and index value of enhancement layer current coded unit, video content is found in Primary layer corresponding
Region, obtains the optimum prediction mode in the region；
5.3) according to the optimum prediction mode of Primary layer, with rough candidate modes, enhancement-layer video is compiled
Code；
5.4) code check and brightness peak signal to noise ratio of seven video sequences are recorded respectively；
Step 6：50 frames before seven video sequences in table 2 are encoded with Yin methods, recording seven respectively regards
The code check and brightness peak signal to noise ratio of frequency sequence.
Step 7：Under the conditions of other seven groups of quantization steps in table 1, repeat step 1.2) to step 6, respectively obtain seven
The code check and brightness peak signal to noise ratio of individual video sequence.
Step 8：Code check and brightness peak signal to noise ratio in step 5 to step 7, according to Video coding joint group
Description in the VCEG-AE07 meeting documents that JCT-VC is proposed, by two groups of code checks of typing and brightness peak signal-to-noise ratio data, and
Loading macro document, obtains seven distortion performance estimator BD-PSNR of video sequence, and to the rate mistake of this seven video sequences
True performance evaluator BD-PSNR averages.
Step 9：According to average distortion performance estimator BD-PSNR, the rough candidate modes of adjustment enhancement layer.
9.1) result according to step 7 judges whether rough candidate modes are final candidate modes：If
Meet -0.055dB ﹤ BD-PSNR ﹤ -0.045dB, then rough candidate modes are exactly final candidate modes, no
Then, rough candidate modes are adjusted, perform step 9.2)；
If 9.2) distortion performance estimator BD-PSNR ﹤ -0.055dB, other of non-rough candidate modes are chosen
Probability highest pattern, used as the rough candidate modes of one of which, but candidate modes number is no more than 7 kinds；
If distortion performance estimator BD-PSNR ﹥ -0.045dB, remove the pattern of probability minimum in rough candidate modes, subtract
Few rough candidate modes number, meets its BD-PSNR and requires, and using adjustment after rough candidate modes as
Final candidate modes, the final candidate modes for obtaining are as shown in 4~table of table 11：
Enhancement layer final candidate modes during table 4 QP0=(20~25) QP1=(QP0+0~QP0+1)
Enhancement layer final candidate modes during table 5 QP0=(20~25) QP1=(QP0+2~QP0+3)
Enhancement layer final candidate modes during table 6 QP0=(26~29) QP1=(QP0+0~QP0+1)
Enhancement layer final candidate modes during table 7 QP0=(26~29) QP1=(QP0+2~QP0+3)
Enhancement layer final candidate modes during table 8 QP0=(30~33) QP1=(QP0+0~QP0+1)
Enhancement layer final candidate modes during table 9 QP0=(30~33) QP1=(QP0+2~QP0+3)
Enhancement layer final candidate modes during table 10 QP0=(34~40) QP1=(QP0+0~QP0+1)
Enhancement layer final candidate modes during table 11 QP0=(34~40) QP1=(QP0+2~QP0+3)
9.3) when video sequence configuration condition is low time delay P, performs step one and arrive step 9.2), obtain final candidate prediction
Pattern is as shown in 12~table of table 19：
Enhancement layer final candidate modes during table 12 QP0=(20~25) QP1=(QP0+0~QP0+1)
Enhancement layer final candidate modes during table 13 QP0=(20~25) QP1=(QP0+2~QP0+3)
Enhancement layer final candidate modes during table 14 QP0=(26~29) QP1=(QP0+0~QP0+1)
Enhancement layer final candidate modes during table 15 QP0=(26~29) QP1=(QP0+2~QP0+3)
Enhancement layer final candidate modes during table 16 QP0=(30~33) QP1=(QP0+0~QP0+1)
Enhancement layer final candidate modes during table 17 QP0=(30~33) QP1=(QP0+2~QP0+3)
Enhancement layer final candidate modes during table 18 QP0=(34~40) QP1=(QP0+0~QP0+1)
Enhancement layer final candidate modes during table 19 QP0=(34~40) QP1=(QP0+2~QP0+3)
9.4) when video sequence configuration condition is Stochastic accessing, performs step one and arrive step 9.2), obtain final candidate pre-
Survey pattern is as shown in 20~table of table 27.
Enhancement layer final candidate modes during table 20 QP0=(20~25) QP1=(QP0+0~QP0+1)
Enhancement layer final candidate modes during table 21 QP0=(20~25) QP1=(QP0+2~QP0+3)
Enhancement layer final candidate modes during table 22 QP0=(26~30) QP1=(QP0+0~QP0+1)
Enhancement layer final candidate modes during table 23 QP0=(26~30) QP1=(QP0+2~QP0+3)
Enhancement layer final candidate modes during table 24 QP0=(31~34) QP1=(QP0+0~QP0+1)
Enhancement layer final candidate modes during table 25 QP0=(31~34) QP1=(QP0+2~QP0+3)
Enhancement layer final candidate modes during table 26 QP0=(35~40) QP1=(QP0+0~QP0+1)
Enhancement layer final candidate modes during table 27 QP0=(35~40) QP1=(QP0+2~QP0+3)
In 4~table of table 27, QP0 represents the value of the quantization step QP of Primary layer, and QP1 represents the quantization step QP's of enhancement layer
Value, 0~pattern of pattern 10 represent respectively SKIP, Inter_2N × 2N, Inter_2N × N, Inter_N × 2N, Inter_2N ×
This 11 kinds of nU, Inter_2N × nD, Inter_nL × 2N, Inter_nR × 2N, Intra_2N × 2N, Intra_N × N and ILR
Predictive mode.
Step 10：According to final candidate modes, each video sequence in table 2 is encoded.
10.1) video sequence configuration condition is low time delay B, quantization step QP0=(20~25), QP1=(QP0+0~QP0
+ 1) when, input video sequence simultaneously carries out down-sampling, obtains Primary layer and enhancement-layer video；
10.2) the first frame of Primary layer and enhancement layer is encoded using Yin methods；
10.3) video to Primary layer and enhancement layer since the second frame is encoded with different methods respectively：If base
This layer is non-I frames, then basic layer video is encoded using efficient video coding HEVC methods, records all coding units
Optimum prediction mode, and the final candidate modes of enhancement layer are found according to Primary layer optimum prediction mode, enhancement layer is regarded
Frequency is encoded；If Primary layer is I frames, step 10.4 is performed)；
10.4) configuration condition of video sequence is judged：If the configuration condition of video sequence is low time delay, using efficient
Video coding HEVC methods are encoded to basic layer video, record the optimum prediction mode of all coding units of Primary layer, root
The final candidate modes of enhancement layer are found according to Primary layer optimum prediction mode, enhancement-layer video is encoded；If depending on
The configuration condition of frequency sequence is Stochastic accessing, then basic layer video is encoded using efficient video coding HEVC methods, profit
Enhancement-layer video is encoded with Yin methods, enhancement layer coding time and whole video sequence coding time are recorded respectively；
10.5) to other seven groups of quantization steps in table 1, step 10.1 is performed respectively) to step 10.4)；
10.6) it is two kinds of situations of low time delay P and Stochastic accessing to video sequence configuration condition, step 10.1 is performed respectively)
To step 10.5), result of the scramble time for obtaining compared with existing Yin methods, as shown in table 28.
The distortion performance and time decrement of the relative Yin methods of the present invention of table 28
In table 28, BD-PSNR represents distortion performance estimator, and unit is dB, and Δ Time represents the present invention and Yin methods
The time change that compares of time, EL represents that the scramble time of enhancement layer compares, when Total represents the coding of whole video sequence
Between compare.
As can be seen from Table 28, when the configuration condition of video sequence is low time delay B, in distortion performance estimator BD-
In the case that PSNR averagely reduces 0.066dB, the enhancement layer coding time reduces 37.77%, and the whole process scramble time is reduced
30.14%；When the configuration condition of frequency sequence is low time delay P, 0.059dB is averagely reduced in distortion performance estimator BD-PSNR
In the case of, the enhancement layer coding time reduces 41.22%, and the whole process scramble time reduces 32.98%；The configuration bar of frequency sequence
When part is Stochastic accessing, in the case where distortion performance estimator BD-PSNR averagely reduces 0.052dB, during enhancement layer coding
Between reduce 45.90%, the whole process scramble time reduces 36.58%.Wherein, the video compress under the configuration condition of Stochastic accessing
Distortion performance preferably, cataloged procedure speed-raising is most fast, illustrates under Stochastic accessing configuration condition, the phase of Primary layer and enhancement layer
Closing property is maximum；And under the configuration condition of low time delay B, the correlation of Primary layer and enhancement layer is minimum.
In sum, the present invention selects the final candidate of enhancement layer pre- using Primary layer and the correlation of enhancement-layer video
Survey pattern, encodes to enhancement-layer video, and finally the scramble time of the fast method makes comparisons with Yin methods.By reality
Test it is concluded that, when the excursion of average BD-PSNR is -0.066dB~-0.052dB, the fast method make enhancement layer put down
The equal scramble time reduces 37.77%~45.90%, and the ensemble average scramble time of video reduces 30.14%~36.58%.Cause
This, under three kinds of video sequence configuration conditions, the present invention can be effectively reduced coding on the basis of distortion performance is ensured
Complexity, improves code rate, can be used for real-time video application.
Foregoing description is preferred embodiment of the invention, it is clear that researcher in this field refers to preferred embodiment of the invention
Various modifications and replacement are made to the present invention with accompanying drawing, these modifications and replacement should all fall under the scope of the present invention.
Claims (3)
1. a kind of SHVC enhancement-layer video coding methods quickly selected based on predictive mode, are comprised the following steps：
(1) corresponding relation of Primary layer coding unit and enhancement layer coding unit is determined：
1.1) in the spatial domain of scalable efficient video coding SHVC is scalable, the quantization step of Primary layer and enhancement layer is given
Value, is input into a video sequence and carries out down-sampling, obtains two groups of different videos of resolution ratio, wherein, the small video of resolution ratio
It is Primary layer, the big video of resolution ratio is enhancement layer；
1.2) the big coding unit of a 2N × 2N of Primary layer is divided into 4 lower Item units of N × N, this 4 N × N's
Coding unit corresponds to 4 coding units of adjacent 2N × 2N of enhancement layer, N=4,8,16,32 respectively；
(2) the Primary layer optimum prediction mode of n frames and the probability of enhancement layer optimum prediction mode are closed before one video sequence of statistics
System, 40 ﹤ n ﹤ 60：
2.1) it is by size 4 × 4 subregions by N × N coding units encoded in Primary layer, there is one kind in each 4 × 4 region
Predictive mode, counts the most predictive mode of access times in these 4 × 4 regions, and compile the pattern as Primary layer N × N
The optimum prediction mode of code unit；
2.2) optimum prediction mode of N × N coding units of Primary layer and corresponding enhancing in n frames before one video sequence of record
The optimum prediction mode of layer 2N × 2N coding units, obtains the general of Primary layer optimum prediction mode and enhancement layer optimum prediction mode
Rate relation；
(3) repeat step (1) and step (2), obtain the Primary layer optimum prediction mode and enhancement layer of multiple video sequences respectively
The probabilistic relation of optimum prediction mode, and probabilistic relation to all video sequences does averagely, obtains predictive mode probability statistics
Table；
(4) in predictive mode probability statistics table, 11 kinds of enhancement layer predictive modes of every kind of Primary layer optimum prediction mode correspondence are right
In every kind of Primary layer optimum prediction mode, the order by its corresponding 11 kinds of enhancement layer predictive mode according to probability from big to small is arranged
Row, it is no directly as rough candidate modes if probability is more than 5% predictive mode and number is no more than 7 kinds
Then, select probability preceding 7 kinds of candidate patterns high are used as rough candidate modes；
(5) the preceding n frames of video sequence are encoded with rough candidate modes, records code check and brightness peak noise
Than；
(6) the preceding n frames of video are encoded with Yin methods, records code check and brightness peak signal to noise ratio；Yin methods are specifically interior
Hold as follows：
(6.1) cycle tests Primary layer be I frames, enhancement layer be P frames under conditions of, have two kinds of systems of selection；First method
It is：Intra patterns and ILR patterns are only selected, other all inter-frame modes are not selected, and second method is：Only selection is used
Inter_2N × 2N patterns of merge, intra patterns and ILR patterns；
(6.2) it is that I frames, enhancement layer are under conditions of B frames or Primary layer are non-I frames, relative to basic in cycle tests Primary layer
For layer prediction process, enhancement layer prediction has two changes：One is in Inter_2N × 2N patterns, to be excluded by estimation
The selection of inter-layer reference image, two is that ILR patterns are done after intra patterns；
(7) the quantization step value of Primary layer and enhancement layer, repeat step (1) to step (6) are changed；
(8) according to step (5) to code check and brightness peak signal to noise ratio in step (7), distortion performance estimator BD- is obtained
PSNR；
(9) result according to step (8) judges whether rough candidate modes are final candidate modes：If full
Foot -0.055dB ﹤ BD-PSNR ﹤ -0.045dB, then rough candidate modes are exactly final candidate modes, otherwise,
Rough candidate modes are adjusted, make its BD-PSNR meet require, and by adjustment after rough candidate modes
As final candidate modes；
(10) according to final candidate modes, video sequence is encoded：
10.1) it is input into a video sequence and carries out down-sampling, obtains Primary layer and enhancement-layer video；
10.2) the first frame of Primary layer and enhancement layer is encoded using Yin methods；
10.3) video to Primary layer and enhancement layer since the second frame is encoded with different methods respectively：If Primary layer
It is non-I frames, then basic layer video is encoded using efficient video coding HEVC methods, records the optimal of all coding units
Predictive mode, and the final candidate modes of enhancement layer are found according to Primary layer optimum prediction mode, enhancement-layer video is entered
Row coding；If Primary layer is I frames, step 10.4 is performed)；
10.4) configuration condition of video sequence is judged：If the configuration condition of video sequence is low time delay, using efficient video
Coding HEVC methods are encoded to basic layer video, the optimum prediction mode of all coding units of Primary layer are recorded, according to base
This layer of optimum prediction mode finds the final candidate modes of enhancement layer, and enhancement-layer video is encoded；If video sequence
The configuration condition of row is Stochastic accessing, then basic layer video is encoded using efficient video coding HEVC methods, using Yin
Method is encoded to enhancement-layer video.
2. the SHVC enhancement-layer video coding methods quickly selected based on predictive mode according to claim 1, wherein walking
Suddenly being encoded to the preceding n frames of video sequence using rough candidate modes described in (5), is carried out as follows：
Basic layer video is encoded with efficient video coding HEVC 2a), the optimum prediction mode of all coding units is recorded；
2b) according to the size and index value of enhancement layer current coded unit, the corresponding region of video content is found in Primary layer,
Obtain the optimum prediction mode in the region；
2c) according to the optimum prediction mode of Primary layer, with rough candidate modes, enhancement-layer video is encoded.
3. the SHVC enhancement-layer video coding methods quickly selected based on predictive mode according to claim 1, wherein walking
Suddenly being adjusted to rough candidate modes described in (9), is by distortion performance estimator BD-PSNR and given threshold
It is compared：
If distortion performance estimator BD-PSNR ﹤ -0.055dB, choose other probability highests of non-rough candidate modes
Pattern, used as the rough candidate modes of one of which, but candidate modes number is no more than 7 kinds；
If distortion performance estimator BD-PSNR ﹥ -0.045dB, remove the mould of probability minimum in rough candidate modes
Formula, reduces rough candidate modes number.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410609900.6A CN104333754B (en) | 2014-11-03 | 2014-11-03 | Based on the SHVC enhancement-layer video coding methods that predictive mode is quickly selected |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410609900.6A CN104333754B (en) | 2014-11-03 | 2014-11-03 | Based on the SHVC enhancement-layer video coding methods that predictive mode is quickly selected |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104333754A CN104333754A (en) | 2015-02-04 |
| CN104333754B true CN104333754B (en) | 2017-06-13 |
Family
ID=52408393
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410609900.6A Expired - Fee Related CN104333754B (en) | 2014-11-03 | 2014-11-03 | Based on the SHVC enhancement-layer video coding methods that predictive mode is quickly selected |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104333754B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109040753A (en) * | 2018-08-17 | 2018-12-18 | 广州酷狗计算机科技有限公司 | Predicting mode selecting method, device and storage medium |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104853188A (en) * | 2015-03-12 | 2015-08-19 | 康佳集团股份有限公司 | Selection control method and system for SVC rapid coding mode in wireless network |
| CN107948649B (en) * | 2016-10-12 | 2020-07-03 | 北京金山云网络技术有限公司 | Video coding method and device based on subjective quality model |
| CN109120926B (en) | 2017-06-23 | 2019-08-13 | 腾讯科技（深圳）有限公司 | Predicting mode selecting method, device and medium |
| CN108235026B (en) * | 2018-02-01 | 2021-09-28 | 重庆邮电大学 | Space-scalable quick coding method |
| CN108259898B (en) * | 2018-02-01 | 2021-09-28 | 重庆邮电大学 | Intra-frame fast coding method based on quality scalable video coding QSHVC |
| CN113225553B (en) * | 2021-04-18 | 2022-09-06 | 南京理工大学 | Method for predicting optimal threshold point in high dynamic video double-layer backward compatible coding system |
| CN113709492B (en) * | 2021-08-25 | 2023-03-24 | 重庆邮电大学 | SHVC (scalable video coding) spatial scalable video coding method based on distribution characteristics |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101562739B (en) * | 2009-06-02 | 2011-07-27 | 北京大学 | Video coding processing method and video coding processing device |
| CN102025995B (en) * | 2010-12-10 | 2012-10-31 | 浙江大学 | Spatial enhancement layer rapid mode selection method of scalable video coding |
| US9307256B2 (en) * | 2013-01-21 | 2016-04-05 | The Regents Of The University Of California | Method and apparatus for spatially scalable video compression and transmission |
| CN103402085A (en) * | 2013-07-19 | 2013-11-20 | 哈尔滨工业大学深圳研究生院 | Mode selection method and system based on video scalable coding compression |
-
2014
- 2014-11-03 CN CN201410609900.6A patent/CN104333754B/en not_active Expired - Fee Related
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109040753A (en) * | 2018-08-17 | 2018-12-18 | 广州酷狗计算机科技有限公司 | Predicting mode selecting method, device and storage medium |
| CN109040753B (en) * | 2018-08-17 | 2020-12-22 | 广州酷狗计算机科技有限公司 | Prediction mode selection method, device and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104333754A (en) | 2015-02-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104333754B (en) | Based on the SHVC enhancement-layer video coding methods that predictive mode is quickly selected | |
| CN107197256B (en) | Method and apparatus for encoding and decoding a sequence of images | |
| US9967587B2 (en) | Apparatus and method for encoding/decoding images | |
| KR102304446B1 (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 | |
| CN101964906B (en) | Rapid intra-frame prediction method and device based on texture characteristics | |
| CN100417228C (en) | Method of selecting in frame prediction mode based on H.264/AVC standard frame image | |
| CN104811718A (en) | Adaptive intra-prediction encoding and decoding method | |
| CN1323555C (en) | Quick selecting method for H.264/AVC multiple reference frame | |
| CN104602003A (en) | Apparatus of generating prediction block | |
| CN1784015A (en) | Inage predicting encoding method in frame | |
| CN103067704B (en) | A kind of method for video coding of skipping in advance based on coding unit level and system | |
| CN104394409B (en) | HEVC predictive mode fast selecting methods based on spatial correlation | |
| CN103596004A (en) | Intra-frame prediction method and device based on mathematical statistics and classification training in HEVC | |
| CN103647975B (en) | Improved sample adaptive offset filtering method based on histogram analysis | |
| CN106937116A (en) | Low-complexity video coding method based on random training set adaptive learning | |
| CN100502507C (en) | Video data compaction coding method | |
| Schroeder et al. | Block structure reuse for multi-rate high efficiency video coding | |
| CN114189688B (en) | Chrominance component prediction method based on luminance template matching | |
| CN1194544C (en) | Video encoding method based on prediction time and space domain conerent movement vectors | |
| CN110913232B (en) | Selection method and device of TU division mode and readable storage medium | |
| CN110139098B (en) | Decision tree-based intra-frame fast algorithm selection method for high-efficiency video encoder | |
| CN100586185C (en) | Mode selection method for transcoding 264 video to reduce resolving capability | |
| CN107343199B (en) | Rapid adaptive compensation method for sampling points in HEVC (high efficiency video coding) | |
| CN113163199B (en) | H265-based video rapid prediction method, rapid coding method and system | |
| CN112243123B (en) | HEVC (high efficiency video coding) rate optimization method based on lossless coding |
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 | ||
| 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: 20170613 |