CN104333756B - HEVC predictive mode fast selecting methods based on relativity of time domain - Google Patents
HEVC predictive mode fast selecting methods based on relativity of time domain Download PDFInfo
- Publication number
- CN104333756B CN104333756B CN201410662687.5A CN201410662687A CN104333756B CN 104333756 B CN104333756 B CN 104333756B CN 201410662687 A CN201410662687 A CN 201410662687A CN 104333756 B CN104333756 B CN 104333756B
- Authority
- CN
- China
- Prior art keywords
- motion
- time domain
- optimum prediction
- prediction mode
- unit
- 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 59
- 230000033001 locomotion Effects 0.000 claims abstract description 80
- 238000013139 quantization Methods 0.000 claims description 21
- 238000002474 experimental method Methods 0.000 claims description 7
- 230000008569 process Effects 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000006835 compression Effects 0.000 description 2
- 238000007906 compression Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention discloses a kind of HEVC predictive mode fast selecting methods based on relativity of time domain, predictive mode selects the problem of computation complexity is high during mainly solving high-effect video encoding standard HEVC interframe encodes.Implementation step is：(1) exercise intensity of video sequence is divided into motion is slow, it is moderate to move, move fast three kinds of states；(2) probabilistic relation of the optimum prediction mode of the optimum prediction mode of coding unit and its time domain adjacent encoder unit when counting slow motion, both moderate exercise intensities of motion respectively；(3) according to probabilistic relation, candidate modes table is built；(4) according to candidate modes table, interframe encode unit is encoded, optimum prediction mode is obtained.The present invention significantly reduces predictive mode selection computation complexity on the premise of video distortion performance is approximately constant, improves predictive mode selection speed, the scramble time is reduced, available for real-time video compress.
Description
Technical field
The invention belongs to field of video processing, interframe encode unit candidate is pre- in more particularly to high-performance video coding HEVC
The generation of survey pattern, available for HEVC interframe encode processes.
Background technology
At present, essentially all high-quality multimedia video application, such as digital video, video request program and DVD, also make
With 2003 formulate Advanced Video Coding standard H.264/AVC.But the development advanced by leaps and bounds with Video Applications, various intelligence
Can terminal, such as smart mobile phone, tablet personal computer have become the instrument of public recreation, for ultra high-definition video demand not yet
Disconnected to increase, people are for high-quality, high-resolution video requirement more and more higher.In order to meet the ever-increasing need of Video Applications
Ask, video encoding standard of new generation is arisen at the historic moment.Video coding joint group JCT-VC started to formulate height of new generation in 2010
Video encoding standard HEVC is imitated, and formal formulate completed in 2 months 2013.HEVC is meant to ensure that the constant feelings of Subjective video quality
Under condition, compression efficiency is doubled, and greatly reduces transmission video signal bandwidth.
HEVC coding frameworks still continue to use the hybrid encoding frame of H.26x series standard, use the block-based mixed of classics
Close encoding model, Union Movement compensation prediction, transition coding and efficient entropy code.With video encoding standard phase before
Than HEVC coding units use flexible quad-tree structure, and size can expeditiously be encoded, predicted from 64 × 64 to 8 × 8
With conversion large-size images block.The experimental configuration condition of HEVC standard includes three kinds：Full I frames configuration, low time delay are configured and random
Access configuration.When Stochastic accessing is configured, using classification B frame, 4 time domain layerings are had.
Predictive mode selection is HEVC key technology, and the purpose is to pre- from multiple candidates to each interframe encode unit
Optimum prediction mode is selected in survey pattern, to obtain optimal precision of prediction.HEVC is carried out optimal using rate-distortion optimization model
The selection of predictive mode.
The predicting mode selecting method of interframe encode unit will be first to SKIP, Inter 2N × 2N, Inter in HEVC standard
2N×N,Inter N×2N,Inter 2N×nU,Inter 2N×nD,Inter nL×2N,Inter nR×2N,Intra
This 10 kinds of candidate modes of 2N × 2N, Intra N × N are traveled through, then using the minimum predictive mode of rate distortion costs as
The optimum prediction mode of current coded unit.On the process of this optimum prediction mode selection causes HEVC encoder complexities drastically
Rise, the real-time application to HEVC brings extreme difficulties.Therefore need on the premise of ensureing that coding efficiency is basically unchanged, reduction
HEVC encoder complexities.
So far, it has been suggested that HEVC predictive mode fast selecting methods have following several：
Motion JCTVC-F045 " Early termination of CU encoding to reduce HEVC
Complexity ", carries out simplifying judgement by coded block flag CBF value to predictive mode.This method is proposed except Inter
During the coded block flag CBF=0 of other inter-frame forecast modes outside N × N, the selection course of remaining predictive mode will be by end
Only, no longer calculated, so as to reduce encoder complexity.This method is referred to as coded block flag quick mode CFM methods.
" Early SKIP Detection for HEVC " are proposed motion JCTVC-G543, first before detection SKIP patterns
Searching I nter 2N × 2N patterns, then detect the motion vector difference DMV and coded block flag of Inter 2N × 2N patterns
CBF, if DMV is equal to (0,0), and CBF is equal to 0, then the optimum prediction mode of current coded unit is just set as SKIP patterns,
No longer remaining predictive mode is traveled through, so as to greatly reduce encoder complexity.This method is referred to as early stage SKIP detections
ESD method.
Jong-Hyeok Lee et al. are in paper " Novel Fast PU Decision Algorithm for the
A kind of fast method based on time-space domain and coding depth correlation is proposed in HEVC Video Standard ".This method is utilized
Coding unit is divided into different motion region by motion complexity：If current coded unit motion complexity is less than threshold value Th1,
Current coded unit is judged for slow moving region, and only SKIP and Inter 2N × 2N patterns are traveled through；If present encoding
Unit motion complexity is more than threshold value Th1 and less than threshold value Th2, then judges current coded unit for moderate moving region, right
SKIP, Inter 2N × 2N, Inter 2N × N and Inter N × 2N patterns are traveled through, and otherwise judge that current coded unit is
10 kinds of predictive modes are all traveled through, have thus skipped the predictive mode of redundancy, effectively reduce coding by fast moving region
Complexity.
Above method reduces HEVC encoder complexity to a certain extent.But, these current methods all do not have
Relativity of time domain is analyzed in detail so that HEVC predictive modes selection speed fails further to be lifted.
The content of the invention
It is an object of the invention to the deficiency for above-mentioned prior art, propose that a kind of HEVC based on relativity of time domain is pre-
Mode quick selecting method is surveyed, in the case where ensureing that video compression performance is basically unchanged, further to reduce encoder complexity,
Lift predictive mode selection speed.
Realizing the object of the invention technological thought is：In predictive mode selection course, according to time domain adjacent encoder unit
Optimum prediction mode, provides the candidate modes of current coded unit, skips redundant prediction pattern as much as possible, Ran Houcong
Optimum prediction mode is selected in candidate modes, coding rate is improved, implementation step includes as follows：
(1) exercise intensity of video sequence is divided into：Motion is slow, move these three fast states of moderate, motion；
(2) determine that the optimum prediction mode of coding unit is related to the optimum prediction mode of its time domain adjacent encoder unit
Property：
2.1) input video sequence and experimental configuration condition, since the 2nd non-I frame, to each interframe encode unit,
The coding unit of same position is searched in reference frame using the numbering of this coding unit, time domain adjacent encoder list is obtained
Member；
2.2) classification of judgment experiment configuration condition：
If experimental configuration condition is low time delay, to [20,26], [27,31], [32,36], [37,41] these four differences
Quantization parameter scope, statistics motion is slow respectively, both moderate exercise intensities of motion coding units optimum prediction moulds
Formula and the probabilistic relation of the optimum prediction mode of its time domain adjacent encoder unit；
If experimental configuration condition is Stochastic accessing, in 1,2,3,4 this four time domains layering, respectively to [20,26],
[27,31], [32,36], [37,41] these four different quantization parameter scopes, statistics motion is slow, move both moderate fortune
The probabilistic relation of the optimum prediction mode of the coding unit of fatigue resistance and the optimum prediction mode of its time domain adjacent encoder unit；
(3) coding unit in the case of these is layered according to experimental configuration condition, quantization parameter scope, exercise intensity and time domain
Optimum prediction mode and its time domain adjacent encoder unit optimum prediction mode probabilistic relation, build low time delay configuration candidate pre-
Survey pattern table and Stochastic accessing configuration candidate modes table：
3.1) optimum prediction mode of selection time domain adjacent encoder unit is SKIP, Inter 2N × 2N, Inter respectively
2N×N,Inter N×2N,Inter 2N×nU,Inter 2N×nD,Inter nL×2N,Inter nR×2N,Intra
Each in this 10 kinds of predictive modes of 2N × 2N, Intra N × N, the optimum prediction mode for calculating current coded unit is this
The probability of 10 kinds of predictive modes, and sorted from big to small by probability；
3.2) since the predictive mode of maximum probability, select probability sum is not less than the prediction mould of 90% and minimum number
Formula sets candidate modes sum no more than 5 kinds as candidate modes, so as to build candidate modes table；
(4) according to candidate modes table, candidate modes are chosen, and interframe encode unit is encoded, are obtained
Optimum prediction mode：
4.1) input video sequence and experimental configuration condition, using the predicting mode selecting method of HEVC standard to I frames and
The coding unit of 1st non-I frame is encoded, and obtains the optimum prediction mode of each coding unit；
4.2) since the 2nd non-I frame, to each interframe encode unit, the classification of its exercise intensity is judged：
If the exercise intensity of current coded unit is slow for motion, step 4.3 is performed)；If the motion of current coded unit
Intensity is moderate to move, and performs step 4.4)；Otherwise, coding unit is carried out using the predicting mode selecting method of HEVC standard
Coding, obtains optimum prediction mode；
4.3) the slow experimental configuration condition classification of motion is judged：
If experimental configuration condition is low time delay, according to quantization parameter QP and the optimum prediction mould of time domain adjacent encoder unit
Corresponding candidate modes are selected to carry out in formula, the candidate modes table for moving slow coding unit under low time delay configuration
Traversal, obtains optimum prediction mode；
If experimental configuration condition is Stochastic accessing, it is layered according to time domain, quantization parameter QP and time domain adjacent encoder unit
Optimum prediction mode, moved under Stochastic accessing configuration and corresponding wait selected in the candidate modes table of slow coding unit
Select predictive mode to be traveled through, obtain optimum prediction mode；
4.4) the moderate experimental configuration condition classification of motion is judged：
If experimental configuration condition is low time delay, according to quantization parameter QP and the optimum prediction mould of time domain adjacent encoder unit
Corresponding candidate modes are selected to carry out in formula, the candidate modes table for moving moderate coding unit under low time delay configuration
Traversal, obtains optimum prediction mode；
If experimental configuration condition is Stochastic accessing, it is layered according to time domain, quantization parameter QP and time domain adjacent encoder unit
Optimum prediction mode, moved under Stochastic accessing configuration and corresponding wait selected in the candidate modes table of moderate coding unit
Select predictive mode to be traveled through, obtain optimum prediction mode.
The present invention has the following advantages that compared with the conventional method：
(a) present invention is due to according to relativity of time domain, having skipped the predictive mode of redundancy so that predictive mode selection is calculated
Complexity is smaller, reduces the substantial amounts of scramble time；
(b) present invention according to different motion intensity due to when selecting candidate modes, being selected so that pattern is selected
Select result more accurate.
Brief description of the drawings
Fig. 1 is the implementation process figure of the present invention；
Fig. 2 is surrounding's coding unit collection location schematic diagram of current coded unit.
Embodiment
The present invention is described in further detail below in conjunction with drawings and examples.The present embodiment is with the technology of the present invention side
Implemented premised on case, give detailed embodiment and specific operation process, but protection scope of the present invention is not limited to
Following embodiments.
Reference picture 1, step is as follows for of the invention realizing：
Step one：The exercise intensity of video sequence is classified.
1a) to current coded unit CU0Set up motion vector set { mv1,mv2,mv3,mv4,mv5,
Wherein mviIt is current coded unit CU0Surrounding coding unit CUiMotion vector, mvi=(xi,yi), i=
1 ..., 5, xiAnd yiThe respectively abscissa and ordinate of motion vector；CU0It is adjacent that the set of surrounding coding unit includes its left side
Coding unit CU1, top adjacent encoder unit CU2, top right-hand side adjacent encoder unit CU3, upper left side adjacent encoder unit CU4With
Time domain adjacent encoder unit CU5, as shown in Figure 2；
1b) calculate the length l of each motion vector in motion vector set：
l(mvi)=| xi|+|yi|；
1c) in step 1b) the maximum L of length is selected in obtained motion vector length set：
L=max { l (mv1),l(mv2),l(mv3),l(mv4),l(mv5),
Motion slow threshold value L0 1d) is set and moderate threshold value L1 is moved：
1d1) video sequence and experimental configuration condition of the input resolution ratio for 416 × 240；
1d2) since first non-I frame, the classification of judgment experiment configuration condition：
If experimental configuration condition is low time delay, export the motion vector of video sequence, according to the size of motion vector and
Video resolution, and with reference to moving the threshold value T0 of slow state in Zeng methods and moving the threshold value T1 of moderate state, set threshold
Value L0 and L1；
If experimental configuration condition is Stochastic accessing, the motion vector of output video sequence different time domain layering, according to fortune
The size and video resolution of dynamic vector, and with reference to the threshold value T0 that slow state is moved in Zeng methods and move moderate state
Threshold value T1, the threshold value L0 and L1 of setting different time domain layering；
Described Zeng methods are shown in Huanqiang Zeng, Canhui Cai et al. article " A Novel Fast Mode
Decision for the H.264/AVC Based on Local MacroblockMotion Activity”；
1d3) repeat step 1d1) and step 1d2) respectively obtain resolution ratio for 832 × 480,1280 × 720,1920 ×
1080th, the slow threshold value L0 of the motion of 2560 × 1600 video sequence and moderate threshold value L1 of motion, obtains final different resolution
Video sequence moves slow threshold value L0 and moves moderate threshold value L1, such as Tables 1 and 2：
The low time delay of the table 1 configuration motion slow state threshold value L0 and moderate state threshold L1 of motion
| Video sequence resolution ratio | Threshold value L0 | Threshold value L1 |
| 416×240 | 2 | 8 |
| 832×480、1280×720 | 3 | 12 |
| 1920×1080 | 8 | 32 |
The Stochastic accessing of the table 2 configuration motion slow state threshold value L0 and moderate state threshold L1 of motion
1e) it is layered according to experimental configuration condition, video resolution and time domain, the slow threshold value of selection motion from table 1 or table 2
L0 and moderate threshold value L1 is moved, by step 1c) in obtained length maximum L, be compared with threshold value L0 and threshold value L1：
If L ＜ L0, exercise intensity is defined slow for motion；
If L0≤L ＜ L1, it is moderate to move to define exercise intensity；
If L >=L1, exercise intensity is defined fast for motion.
Step 2：Determine the optimum prediction mode of coding unit and the optimum prediction mode of its time domain adjacent encoder unit
Correlation.
2a) input video sequence BQSquare and experimental configuration condition, since the 2nd non-I frame, are compiled to each interframe
Code unit, is searched for the coding unit of same position in reference frame using the numbering of this coding unit, obtains time domain adjacent
Coding unit；
2b) the classification of judgment experiment configuration condition：
If experimental configuration condition be low time delay, test video sequence BQSquare preceding 100 frame, to [20,26], [27,
31], [32,36], [37,41] these four different quantization parameter scopes, statistics motion is slow respectively, move both moderate fortune
The probabilistic relation of the coding unit optimum prediction mode of fatigue resistance and its time domain adjacent encoder unit optimum prediction mode；
If experimental configuration condition be Stochastic accessing, test video sequence BQSquare preceding 65 frame, 1,2,3,4 this four
In individual time domain layering, respectively to [20,26], [27,31], [32,36], [37,41] these four different quantization parameter scopes, system
Meter motion is slow, the optimum prediction mode and its time domain adjacent encoder unit of the coding unit of both moderate exercise intensities of motion
Optimum prediction mode probabilistic relation；
2c) repeat step 2a) and step 2b) respectively obtain video sequence BasketballPass and Johnny coding list
The probabilistic relation of the optimum prediction mode of member and the optimum prediction mode of its time domain adjacent encoder unit, to these three above-mentioned videos
The probabilistic relation of sequence is averaged, and obtains the coding list that final low time delay configuration configures lower different motion intensity with Stochastic accessing
Member and the probabilistic relation of the optimum prediction mode of its time domain adjacent encoder unit, only provide the lower quantization parameter of low time delay configuration below
Probabilistic relation when scope is [20,26], and the lower time domain of Stochastic accessing configuration be layered as 1 and quantization parameter scope for [20,
26] probabilistic relation, such as table 3 are to table 6：
The probability of the low time delay of the table 3 slow coding unit of configuration motion and the optimum prediction mode of time domain adjacent encoder unit
Relation
The probability of the low time delay of the table 4 moderate coding unit of configuration motion and the optimum prediction mode of time domain adjacent encoder unit
Relation
The Stochastic accessing of the table 5 slow coding unit of configuration motion is general with the optimum prediction mode of time domain adjacent encoder unit
Rate relation
The Stochastic accessing of the table 6 moderate coding unit of configuration motion is general with the optimum prediction mode of time domain adjacent encoder unit
Rate relation
In above-mentioned table 3- tables 6, pattern 0 represents SKIP, and pattern 1 represents Inter 2N × 2N, pattern 2 represent Inter 2N ×
N, pattern 3 represents Inter N × 2N, and pattern 4 represents Inter 2N × nU, and pattern 5 represents Inter 2N × nD, and pattern 6 is represented
Inter nL × 2N, mode 7 represents Inter nR × 2N, and pattern 8 represents Intra 2N × 2N, and pattern 9 represents Intra N × N.
Step 3：According to table 3- tables 6, low time delay configuration candidate modes table and Stochastic accessing configuration candidate prediction are built
Pattern table.
3a) optimum prediction mode of coding unit and its time domain adjacent encoder unit includes：SKIP,Inter 2N×2N,
Inter 2N×N,Inter N×2N,Inter 2N×nU,Inter 2N×nD,Inter nL×2N,Inter nR×2N,
This 10 kinds of predictive modes of Intra 2N × 2N, Intra N × N；Selection time domain adjacent encoder unit optimum prediction mode be
SKIP patterns, the optimum prediction mode for calculating current coded unit is respectively the probability of above-mentioned 10 kinds of predictive modes, and by probability
Sort from big to small；
3b) repeat step 3a) select the optimum prediction mode of time domain adjacent encoder unit to predict moulds for other 9 kinds respectively
Formula, the optimum prediction mode for calculating current coded unit is the probability of above-mentioned 10 kinds of predictive modes, and is arranged from big to small by probability
Sequence；
3c) since the predictive mode of maximum probability, select probability sum is not less than the prediction mould of 90% and minimum number
Formula sets candidate modes sum no more than 5 kinds as candidate modes, so as to build candidate modes table, such as
Table 7- tables 10；
The slow coding unit candidate modes table of the low time delay of table 7 configuration motion
The moderate coding unit candidate modes table of the low time delay of table 8 configuration motion
The slow coding unit candidate modes table of the Stochastic accessing of table 9 configuration motion
The moderate coding unit candidate modes table of the Stochastic accessing of table 10 configuration motion
Step 4：According to table 7- tables 10, candidate modes are chosen, to the interframe encode list of each video sequence in table 11
Member is encoded, and obtains optimum prediction mode.
4a) input video sequence, using the predicting mode selecting method of HEVC standard to I frames and the coding of the 1st non-I frame
Unit is encoded, and obtains the optimum prediction mode of each coding unit；
4b) since the 2nd non-I frame, to each interframe encode unit, its exercise intensity is judged：
If the exercise intensity of current coded unit is slow for motion, step 4c is performed)；If the motion of current coded unit is strong
Spend, execution step 4d moderate to move)；Otherwise, coding unit is compiled using the predicting mode selecting method of HEVC standard
Code, obtains optimum prediction mode；
4c) judge the slow experimental configuration condition classification of motion：
If experimental configuration condition is low time delay, according to quantization parameter QP and the optimum prediction mould of time domain adjacent encoder unit
Formula, selects corresponding candidate modes to be traveled through in table 7, obtains optimum prediction mode；
If experimental configuration condition is Stochastic accessing, it is layered according to time domain, quantization parameter QP and time domain adjacent encoder unit
Optimum prediction mode, select corresponding candidate modes to be traveled through in table 9, obtain optimum prediction mode；
4d) judge the moderate experimental configuration condition classification of motion：
If experimental configuration condition is low time delay, according to quantization parameter QP and the optimum prediction mould of time domain adjacent encoder unit
Formula, selects corresponding candidate modes to be traveled through in table 8, obtains optimum prediction mode；
If experimental configuration condition is Stochastic accessing, it is layered according to time domain, quantization parameter QP and time domain adjacent encoder unit
Optimum prediction mode, select corresponding candidate modes to be traveled through in table 10, obtain optimum prediction mode.
The effect of the present invention can be further illustrated by following emulation：
1. experimental situation
Using VS2010 coding environments, tested with reference software HM16.0, experimental configuration condition configures for low time delay
With Stochastic accessing configuration.
The video sequence details of experiment test such as table 11：
The video sequence details of table 11
2. experiment content
Using the inventive method, CFM fast methods, ESD fast methods and Lee fast methods respectively to all in table 11
Video sequence is encoded, record scramble time and distortion performance estimator BD-PSNR.The present invention and these three fast methods
Coding efficiency compare such as table 12- tables 14, wherein table 12 is the comparison of the inventive method and CFM fast method coding efficiencies, table
13 be the comparison of the inventive method and ESD fast method coding efficiencies, and table 14 is that the inventive method and Lee fast methods are Encoding
The comparison of energy.
Wherein, in table 12- tables 14Represent the inventive method and existing quick
Time variation amount is compared in system of selection, and "-" represents that the inventive method raises speed than existing fast method in terms of the time.BD-
PSNR represented under given equal code check, the brightness peak signal to noise ratio PSNR-Y of two methods difference, and its unit is dB,
"-" represents that the inventive method is reduced than existing fast method PSNR-Y.
The inventive method of table 12 is compared with CFM fast methods
From table 12, the inventive method is compared with CFM fast methods, and low time delay is averagely reduced when configuring in BD-PSNR
In the case of 0.07dB, the scramble time has averagely raised speed 4.69%, and Stochastic accessing averagely reduces 0.07dB when configuring in BD-PSNR
In the case of, the scramble time has averagely raised speed 8.43%.
The inventive method of table 13 is compared with ESD fast methods
From table 13, the inventive method is compared with ESD fast methods, and low time delay is averagely reduced when configuring in BD-PSNR
In the case of 0.095dB, the scramble time has averagely raised speed 13.81%, and Stochastic accessing is averagely reduced when configuring in BD-PSNR
In the case of 0.10dB, the scramble time has averagely raised speed 11.68%.
The inventive method of table 14 is compared with Lee fast methods
From table 14, the inventive method is compared with Lee fast methods, and Stochastic accessing averagely drops when configuring in BD-PSNR
In the case of low 0.04dB, the scramble time has averagely raised speed 4.22%.
To sum up, the present invention utilizes relativity of time domain, skips the predictive mode of redundancy, in the essentially identical situations of BD-PSNR
Under, further improve the speed of predictive mode selection.
Foregoing description is preferred embodiment of the invention, it is clear that researcher in this field refers to the preferred embodiment of the present invention
Various modifications and replacement are made to the present invention with accompanying drawing, these modifications and replacement should all be fallen under the scope of the present invention.
Claims (2)
1. a kind of HEVC predictive mode fast selecting methods based on relativity of time domain, it is characterised in that
Comprise the following steps：
(1) exercise intensity of video sequence is divided into：Motion is slow, move these three fast states of moderate, motion；
(2) correlation of the optimum prediction mode and the optimum prediction mode of its time domain adjacent encoder unit of coding unit is determined：
2.1) input video sequence and experimental configuration condition, since the 2nd non-I frame, to each interframe encode unit, are used
The numbering of this coding unit searches for the coding unit of same position in reference frame, obtains time domain adjacent encoder unit；
2.2) classification of judgment experiment configuration condition：
If experimental configuration condition is low time delay, to [20,26], [27,31], [32,36], [37,41] these four different amounts
Change parameter area, statistics motion is slow respectively, both moderate exercise intensities of motion coding units optimum prediction modes with
The probabilistic relation of the optimum prediction mode of its time domain adjacent encoder unit；
If experimental configuration condition is Stochastic accessing, in 1,2,3,4 this four time domains layering, respectively to [20,26], [27,
31], [32,36], [37,41] these four different quantization parameter scopes, statistics motion is slow, it is strong to move both moderate motions
The probabilistic relation of the optimum prediction mode of the coding unit of degree and the optimum prediction mode of its time domain adjacent encoder unit；
(3) coding unit is most in the case of being layered these according to experimental configuration condition, quantization parameter scope, exercise intensity and time domain
The probabilistic relation of good predictive mode and its time domain adjacent encoder unit optimum prediction mode, builds low time delay configuration candidate prediction mould
Formula table and Stochastic accessing configuration candidate modes table：
3.1) optimum prediction mode of time domain adjacent encoder unit is selected respectively for SKIP, Inter 2N × 2N, Inter 2N ×
N,Inter N×2N,Inter 2N×nU,Inter 2N×nD,Inter nL×2N,Inter nR×2N,Intra 2N×
Each in this 10 kinds of predictive modes of 2N, Intra N × N, the optimum prediction mode for calculating current coded unit is this 10 kinds
The probability of predictive mode, and sorted from big to small by probability；
3.2) since the predictive mode of maximum probability, select probability sum is not less than the predictive mode work of 90% and minimum number
For candidate modes, and set candidate modes sum and be no more than 5 kinds, so as to build candidate modes table；
(4) according to candidate modes table, candidate modes are chosen, and interframe encode unit is encoded, obtain optimal
Predictive mode：
4.1) input video sequence and experimental configuration condition, using the predicting mode selecting method of HEVC standard to I frames and the 1st
The coding unit of non-I frames is encoded, and obtains the optimum prediction mode of each coding unit；
4.2) since the 2nd non-I frame, to each interframe encode unit, the classification of its exercise intensity is judged：
If the exercise intensity of current coded unit is slow for motion, step 4.3 is performed)；If the exercise intensity of current coded unit
It is moderate to move, perform step 4.4)；Otherwise, coding unit is compiled using the predicting mode selecting method of HEVC standard
Code, obtains optimum prediction mode；
4.3) the slow experimental configuration condition classification of motion is judged：
If experimental configuration condition is low time delay, according to quantization parameter QP and the optimum prediction mode of time domain adjacent encoder unit,
Corresponding candidate modes progress time is selected in the candidate modes table for moving slow coding unit under low time delay configuration
Go through, obtain optimum prediction mode；
If experimental configuration condition is Stochastic accessing, be layered according to time domain, quantization parameter QP and time domain adjacent encoder unit most
Select corresponding candidate pre- in good predictive mode, the candidate modes table for moving slow coding unit under Stochastic accessing configuration
Survey pattern is traveled through, and obtains optimum prediction mode；
4.4) the moderate experimental configuration condition classification of motion is judged：
If experimental configuration condition is low time delay, according to quantization parameter QP and the optimum prediction mode of time domain adjacent encoder unit,
Corresponding candidate modes progress time is selected in the candidate modes table for moving moderate coding unit under low time delay configuration
Go through, obtain optimum prediction mode；
If experimental configuration condition is Stochastic accessing, be layered according to time domain, quantization parameter QP and time domain adjacent encoder unit most
Select corresponding candidate pre- in good predictive mode, the candidate modes table for moving moderate coding unit under Stochastic accessing configuration
Survey pattern is traveled through, and obtains optimum prediction mode.
2. the HEVC predictive mode fast selecting methods as claimed in claim 1 based on relativity of time domain, it is characterised in that：Institute
State and be divided into the exercise intensity of video sequence in step (1)：Motion is slow, these three fast states of moderate, motion are moved, by as follows
Step is carried out：
1a) to current coded unit CU0Set up motion vector set { mv1,mv2,mv3,mv4,mv5,
Wherein mviIt is current coded unit CU0Surrounding coding unit CUiMotion vector, mvi=(xi,yi), i=1 ..., 5, xi
And yiThe respectively abscissa and ordinate of motion vector；CU0The coding unit of surrounding includes its left side adjacent encoder unit CU1、
Top adjacent encoder unit CU2, top right-hand side adjacent encoder unit CU3, upper left side adjacent encoder unit CU4With time domain adjacent encoder
Unit CU5；
1b) calculate the length l of each motion vector in motion vector set：
l(mvi)=| xi|+|yi|,
1c) in step 1b) the maximum L of length is selected in obtained motion vector length set：
L=max { l (mv1),l(mv2),l(mv3),l(mv4),l(mv5)}；
Motion slow threshold value L0 1d) is set and moderate threshold value L1 is moved：
1d1) video sequence and experimental configuration condition of the input resolution ratio for 416 × 240；
1d2) since first non-I frame, the classification of judgment experiment configuration condition：
1d3) repeat step 1d1) and step 1d2) respectively obtain resolution ratio for 832 × 480,1280 × 720,1920 × 1080,
The slow threshold value L0 of motion and the moderate threshold value L1 of motion of 2560 × 1600 video sequence, obtain final different resolution video sequence
The slow threshold value L0 of the row motion and moderate threshold value L1 of motion；
1e) it is layered according to experimental configuration condition, video resolution and time domain, the slow threshold value L0 of selection motion and the moderate threshold value of motion
L1, by step 1c) in obtained length maximum L, be compared with threshold value L0 and threshold value L1：
If L ＜ L0, exercise intensity is defined slow for motion；
If L0≤L ＜ L1, it is moderate to move to define exercise intensity；
If L >=L1, exercise intensity motion is defined fast.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410662687.5A CN104333756B (en) | 2014-11-19 | 2014-11-19 | HEVC predictive mode fast selecting methods based on relativity of time domain |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410662687.5A CN104333756B (en) | 2014-11-19 | 2014-11-19 | HEVC predictive mode fast selecting methods based on relativity of time domain |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104333756A CN104333756A (en) | 2015-02-04 |
| CN104333756B true CN104333756B (en) | 2017-10-24 |
Family
ID=52408395
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410662687.5A Expired - Fee Related CN104333756B (en) | 2014-11-19 | 2014-11-19 | HEVC predictive mode fast selecting methods based on relativity of time domain |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104333756B (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104902271B (en) * | 2015-05-15 | 2017-12-29 | 腾讯科技（北京）有限公司 | Predicting mode selecting method and device |
| CN106331699B (en) * | 2015-06-19 | 2019-04-26 | 华为技术有限公司 | Image encoding method and image encoder |
| CN106170089B (en) * | 2016-08-25 | 2020-05-22 | 上海交通大学 | H.265-based multi-path coding method |
| CN107566846B (en) * | 2017-09-20 | 2020-04-14 | 北京字节跳动网络技术有限公司 | Video coding skipping mode decision method, device, equipment and storage medium |
| CN108989801B (en) * | 2018-07-27 | 2021-10-08 | 福州大学至诚学院 | Space-time characteristic-based adaptive mode selection rapid method |
| CN113596457A (en) * | 2019-09-23 | 2021-11-02 | 杭州海康威视数字技术股份有限公司 | Encoding and decoding method, device and equipment |
| CN112839224B (en) * | 2019-11-22 | 2023-10-10 | 腾讯科技（深圳）有限公司 | Prediction mode selection method and device, video coding equipment and storage medium |
| TWI743919B (en) | 2020-08-03 | 2021-10-21 | 緯創資通股份有限公司 | Video processing apparatus and processing method of video stream |
| TWI749676B (en) | 2020-08-03 | 2021-12-11 | 緯創資通股份有限公司 | Image quality assessment apparatus and image quality assessment method thereof |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102984521A (en) * | 2012-12-12 | 2013-03-20 | 四川大学 | High-efficiency video coding inter-frame mode judging method based on temporal relativity |
| US20130336384A1 (en) * | 2012-06-15 | 2013-12-19 | General Instrument Corporation | Method and apparatus for efficient slice header processing |
-
2014
- 2014-11-19 CN CN201410662687.5A patent/CN104333756B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130336384A1 (en) * | 2012-06-15 | 2013-12-19 | General Instrument Corporation | Method and apparatus for efficient slice header processing |
| CN102984521A (en) * | 2012-12-12 | 2013-03-20 | 四川大学 | High-efficiency video coding inter-frame mode judging method based on temporal relativity |
Non-Patent Citations (3)
| Title |
|---|
| A Novel Fast Mode Decision for the H.264/AVC Based on Local Macroblock Motion Activity;Zeng Huanqiang等;《IEEE International Conference on Image and Graphics》;20070824;263-267 * |
| Novel fast PU decision algorithm for the HEVC video standard;LEE J H等;《IEEE International Conference on Image Processing》;20130918;1982-1985 * |
| 高效率视频编码帧内快速算法;许东旭等;《计算机应用》;20140810;第34卷(第8期);2375-2379 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104333756A (en) | 2015-02-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104333756B (en) | HEVC predictive mode fast selecting methods based on relativity of time domain | |
| CN104754357B (en) | Intraframe coding optimization method and device based on convolutional neural networks | |
| CN102984521B (en) | High-efficiency video coding inter-frame mode judging method based on temporal relativity | |
| CN103188496B (en) | Based on the method for coding quick movement estimation video of motion vector distribution prediction | |
| CN100551075C (en) | A kind of method for choosing frame inner forecast mode of low complex degree | |
| CN103327325B (en) | The quick self-adapted system of selection of intra prediction mode based on HEVC standard | |
| CN103546749B (en) | Method for optimizing HEVC (high efficiency video coding) residual coding by using residual coefficient distribution features and bayes theorem | |
| CN102740077B (en) | H.264/AVC standard-based intra-frame prediction mode selection method | |
| CN107147911A (en) | LIC quick interframe coding mode selection method and device is compensated based on local luminance | |
| CN104394409B (en) | HEVC predictive mode fast selecting methods based on spatial correlation | |
| CN102025995B (en) | Spatial enhancement layer rapid mode selection method of scalable video coding | |
| CN103517069A (en) | HEVC intra-frame prediction quick mode selection method based on texture analysis | |
| CN101640802A (en) | Video inter-frame compression coding method based on macroblock features and statistical properties | |
| CN105791826A (en) | Data mining-based HEVC inter-frame fast mode selection method | |
| CN105120290B (en) | A kind of deep video fast encoding method | |
| CN103997645B (en) | Quick HEVC intra-frame coding unit and pattern decision method | |
| CN103384325A (en) | Quick inter-frame prediction mode selection method for AVS-M video coding | |
| CN108366256A (en) | A kind of HEVC intra prediction modes quickly select system and method | |
| CN105898332B (en) | For the fast deep figure frame mode decision method of 3D-HEVC coding standards | |
| Chen et al. | A novel fast intra mode decision for versatile video coding | |
| CN107277506A (en) | A kind of motion vector accuracy fast selecting method and device based on adaptive motion vector precision | |
| CN104853191A (en) | HEVC fast coding method | |
| CN101304529A (en) | Method and device for selecting macro block pattern | |
| CN105657420A (en) | HEVC-oriented fast intra-frame prediction mode decision method and device | |
| CN103533355A (en) | Quick coding method for HEVC (high efficiency video 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 |
Granted publication date: 20171024 |