CN107343198A - A kind of quick decision method of AVS2 inter-frame forecast modes - Google Patents
A kind of quick decision method of AVS2 inter-frame forecast modes Download PDFInfo
- Publication number
- CN107343198A CN107343198A CN201710315131.2A CN201710315131A CN107343198A CN 107343198 A CN107343198 A CN 107343198A CN 201710315131 A CN201710315131 A CN 201710315131A CN 107343198 A CN107343198 A CN 107343198A
- Authority
- CN
- China
- Prior art keywords
- mse
- mode
- mrow
- hor
- ver
- 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.)
- Pending
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
- H04N19/103—Selection of coding mode or of prediction mode
- H04N19/109—Selection of coding mode or of prediction mode among a plurality of temporal predictive coding modes
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
- H04N19/146—Data rate or code amount at the encoder output
- H04N19/147—Data rate or code amount at the encoder output according to rate distortion criteria
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
- H04N19/17—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object
- H04N19/176—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object the region being a block, e.g. a macroblock
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/503—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
- H04N19/51—Motion estimation or motion compensation
- H04N19/563—Motion estimation with padding, i.e. with filling of non-object values in an arbitrarily shaped picture block or region for estimation purposes
Abstract
The problem of present invention is directed to AVS2 interframe encode complexities, proposes a kind of method of fast mode judgment.When carrying out interframe encode, the Texture complication of calculation code block horizontally and vertically first, the symmetrical partition mode of calculated level (HOR_SYM) if the Texture complication of horizontal direction is smaller, otherwise calculate vertical symmetry pattern (VER_SYM).The predictive mode of optimal reference block is used to avoid producing false judgment because of external factor as reference simultaneously.If horizontal symmetrical or vertical symmetry pattern are current best mode, further analysis level is asymmetric or vertical asymmetric mode under predicting unit (PU) texture, and chosen whether according to textural characteristics to carry out the calculating of each asymmetric mode.Such a method can at most reduce by 5 kinds of mode computations, can reduce by 3 kinds of mode computations at least, can significantly reduce encoder complexity, while coding efficiency only has minimum loss.
Description
Technical field
The present invention relates to a kind of quick decision method of AVS2 inter-frame forecast modes, is a kind of time-space domain phase based on encoding block
The algorithm of closing property and texture special type.
Background technology
AVS2 is the autonomous audio/video encoding standard of country of new generation, and its compression efficiency compares previous generation standards (abbreviation AVS1)
About one times is H.264/AVC improved with international standard, especially there is higher performance boost in terms of scene Video coding.AVS2 is adopted
With many advanced coding techniques, most representational is flexible quad-tree partition structure and diversified predictive mode.
Wherein in terms of infra-frame prediction, brightness has 30 directional prediction modes and 3 non-directional prediction modes;In terms of inter prediction then
It with the addition of the inter-frame forecast modes such as double forward predictions.Advanced coding techniques improves code efficiency, but cataloged procedure is more
Add complexity.Such as when being encoded using quad-tree partition structure, if the size of minimum coding unit is 8 × 8, often
Individual 64 × 64 block shares 1+174=83522 kinds of possible dividing modes.And AVS2 employs multi-reference frame technology, interframe
Each encoding block is required for traveling through the reference frame in buffering area to find optimal reference block when prediction, further increases multiple
Miscellaneous degree.
Largely time-consuming calculating has had a strong impact on the applications of AVS2 in practice.AVS2 is in inter predication process, not
There are the textural characteristics for considering sequence, be all that traversal attempts all predictive modes to find optimal mode.In fact, time domain and sky
The textural characteristics of the adjacent encoding block in domain are all very close, and they are also very big using the probability of identical partition mode.And each mould
Also there is certain contact between formula, such as when the rate distortion of horizontal symmetrical pattern is smaller than the rate distortion of vertical symmetry, then hang down
Straight asymmetric mode is unlikely to be optimal mode.The research for this respect also had much in the past, such as the encoded figure of surrounding
As the predictive mode of block is predicted as candidate pattern to the block currently encoded, although these methods are to reducing current block
Mode computation have a certain effect, but often because Information Pull obtains partial birth life error, and error can be next to predicting
Coding unit has an impact.Some algorithms can also set a threshold value and be used to judge whether to predict, the essence of threshold value setting
True property is very crucial, and its accuracy directly affects the effect of model selection.
The content of the invention
It is an object of the invention to provide a kind of quick decision algorithm of AVS2 inter-frame forecast modes, this method utilizes coding unit
(CU, Coding Unit) and predicting unit (PU, Prediction Unit) time domain and the information in spatial domain are single in conjunction with coding
The texture information of member predicts its dividing mode, so as to avoid traveling through all partition modes during interframe encode.
To reach above-mentioned purpose, idea of the invention is that:
Before carrying out model prediction to current CU, multiple time domains, spatial domain adjacent block can be chosen as reference block, and from
One is chosen in reference block with the immediate reference block of present encoding block texture as optimal reference block.Then analyze current CU's
Texture features, judge to carry out horizontal symmetrical division according to texture features or carry out vertical symmetry division, while by optimal reference
The predictive mode of block as a comparison, avoids producing mistake.After horizontal or vertical division is determined, further analyze CU's
Textural characteristics, decide whether to carry out asymmetric division.The complexity that can make inter prediction by the algorithm greatly reduces, and
The presence of optimal reference block ensure that the accuracy of algorithm.
Conceive more than, the technical scheme is that:
A kind of quick decision method of AVS2 inter-frame forecast modes, operating procedure are as follows:
(1) before being encoded to current CU, choose previous frame correspondence position same depth CU, and the same frame left side and
The CU of the same depth in top adjacent position as CU refer to, calculate it is each with reference to CU and current CU Texture complication, and from owning
Reference CU in selection one with current CU textures it is immediate be used as optimal reference block, record optimal reference block partition mode
For Mode0;
(2) rate distortion costs of NO_SPLIT patterns are calculated, and current CU optimum prediction mould BestMode is set to NO_
SPLIT patterns;
(3) the horizontal mean square error MSE of more current CUHorWith vertical mean square error MSEVerSize, if MSEHor<
MSEVer, jump to step (4);If MSEVer<MSEHor, jump to step (5);
(4) if Mode0 is one kind in three kinds of vertical division patterns VER_SYM, VER_LEFT, VER_RIGHT, calculate
Horizontal symmetrical pattern HOR_SYM and vertical symmetry pattern VER_SYM rate distortion costs, and compared with current best mode,
The pattern of Least-cost is selected as optimum prediction mode BestMode;Otherwise the rate of a calculated level symmetric pattern HOR_SYM
Distortion cost, and optimal mode BestMode is relatively obtained with current best mode;
(5) if Mode0 is one kind in three kinds of horizontal division patterns HOR_SYM, HOR_UP, HOR_DOWN, calculate and hang down
Straight symmetric pattern VER_SYM and horizontal symmetrical pattern HOR_SYM rate distortion costs, and compared with current best mode, choosing
The pattern of Least-cost is selected as optimum prediction mode BestMode;Otherwise the rate for only calculating vertical symmetry pattern VER_SYM is lost
True cost, and optimal mode BestMode is relatively obtained with current best mode;
(6) if optimal mode BestMode is NO_SPLIT patterns after step (4) or step (5), after terminating
Continuous asymmetric mode calculates, and jumps to step (9);
(7) if optimal mode BestMode is horizontal symmetrical pattern HOR_SYM after step (4) or step (5),
Calculated level asymmetric mode HOR_UP (HOR_DOWN) top position PU horizontal mean square error MSEUpUp(MSEDownUp) and under
Put PU horizontal mean square error MSE in orientationUpDown(MSEDownDown).If MSEUpUp<MSEUpDown(MSEDownDown<MSEDownUp), then
Mode computation is carried out to horizontal asymmetrical pattern HOR_UP (HOR_DOWN);Otherwise the pattern is skipped.
(8) if optimal mode BestMode is vertical symmetry pattern VER_SYM after step (4) or step (5),
Calculate vertical asymmetric mode VER_LEFT (VER_RIGHT) leftward positions PU vertical mean square error MSELeftLeft
(MSERightLeft) and right positions PU vertical mean square error MSELeftRight(MSERightRight).If MSELeftLeft<
MSELeftRight(MSERightRight<MSERightLeft), then row mode is entered to vertical asymmetric mode VER_LEFT (VER_RIGHT)
Calculate;Otherwise the pattern is skipped.
(9) more other pattern of surplus.
The Texture complication present invention of encoding block is represented using least mean-square error MSE in above-mentioned steps (1).Can profit
The mean square errors of CU both vertically and horizontally are calculated with formula (1) and (2).
In formula, MSEVerAnd MSEHorRepresent the mean square errors of CU both vertically and horizontally;W and H is to be encoded respectively
CU width and height (in units of pixel);α, β are coefficient values, for adjust under different demarcation pattern level with it is vertical
Pixel number on direction, calculate MSEVerAnd MSEHorWhen, α and β value are all 1;P (x, y) represents the value of CU pixels;mxWith
myCalculating MSEVerAnd MSEHorWhen represent respectively current CU (x+1)th row pixel average value and y+1 row pixel
The average value of point.
The similarity of two encoding block textural characteristics, horizontal mean square error MSE can be utilizedHorWith vertical mean square error
MSEVerIt is absolute error and represent, such as formula (3).The textural characteristics that abs is smaller to mean that two encoding blocks are closer.
Abs=| MSEHor1-MSEHor2|+|MSEVer1-MSEVer2| (3)
The calculation formula of rate distortion costs in above-mentioned steps (2) is:
J(s,c,mode|QP,λmode)=SAD (s, c, mode | QP)+λmotionR(s,c,mode|QP) (4)
SAD is represented under conditions of given quantization parameter QP in formula, the predicting unit of the pattern and the difference of reconstruction unit block
It is worth quadratic sum;S is current prediction unit;C is reconstruction image block;Mode is selected interframe encoding mode;R is compiled under the pattern
Bit number needed for code;Lagrange factor λmotionIt is to be obtained according to quantization parameter QP.
The principle entered a judgement in above-mentioned steps (3):If on some direction of image the Texture complication (MSE) of pixel compared with
It is small, that is, fluctuating quantity is smaller, then the pixel on this direction belongs to the probability of same target with regard to larger.
The predictive mode Mode0 that optimal reference block is used in above-mentioned steps (4) and (5) is in order to avoid because external factor
And produce false judgment;Such as due to the influence of other external factor such as light may cause some and be not belonging to same object
Pixel there is pixel value relatively.
Above-mentioned steps (6) if in horizontal symmetrical partition mode HOR_SYM and vertical division pattern VER_SYM rate distortion generation
Valency is all bigger than NO_SPLIT, then horizontal asymmetrical and vertical asymmetric division are also likely to be optimal mode.
Above-mentioned steps (7) if in current best mode be horizontal symmetrical pattern, further analysis level asymmetric mode
Lower predicting unit PU Texture complication, and decided whether according to PU textural characteristics to carry out the calculating of each asymmetric mode,
Rather than all horizontal asymmetrical patterns of traversal;During the asymmetrical both of which HOR_UP and HOR_DOWN of analysis level, use
Formula (2), and use MSEUpUpAnd MSEUpDownThe MSE of the block up and down under HOR_UP patterns is represented respectively, and α value is all 1, β value
Respectively 0.25 and 0.75;Use MSEDownUpAnd MSEDownDownThe MSE of the block up and down under HOR_DOWN patterns, α value are represented respectively
It is respectively 0.75 and 0.25 all for 1, β value;M when calculating this 4 PU MSExRepresent the flat of current PU (x+1)th row pixel
Average.
Above-mentioned steps (8) if in current best mode be vertical symmetry pattern, further analyze vertical asymmetric mode
Lower predicting unit PU Texture complication, and decided whether according to PU textural characteristics to carry out the calculating of each asymmetric mode,
Rather than all vertical asymmetric modes of traversal;When analyzing vertical asymmetrical both of which VER_LEFT and VER_RIGHT, make
With formula (1), and use MSELeftLeftAnd MSELeftRightThe MSE of the left and right block under VER_LEFT patterns, α value point are represented respectively
Not Wei 0.25 and 0.75, β value be all 1;Use MSERightLeftAnd MSERightRightThe left side under VER_RIGHT patterns is represented respectively
Right piece of MSE, α value are respectively 0.75 and 0.25, and β value is all 1;M when calculating this 4 PU MSEyRepresent current PU y+
The average value of 1 row pixel.
Pattern of surplus in above-mentioned steps (9) includes SKIP patterns and SPLIT patterns.
The present invention compared with prior art, there is following obvious prominent substantive distinguishing features and notable technology to enter
Step:
The problem of present invention is directed to AVS2 interframe encode complexities, proposes a kind of method of fast mode judgment.Carry out
During interframe encode, the Texture complication of calculation code block horizontally and vertically first, if the texture of horizontal direction is complicated
The smaller symmetrical partition mode of then calculated level (HOR_SYM) of degree, otherwise calculates vertical symmetry pattern (VER_SYM).Use simultaneously
The predictive mode of optimal reference block avoids producing false judgment because of external factor as reference.If horizontal symmetrical is vertical
To being referred to as current best mode, then further analysis level is asymmetric or vertical asymmetric mode under PU texture, and according to line
Whether reason feature selecting carries out the calculating of each asymmetric mode.Such a method can at most reduce by 5 kinds of mode computations, at least may be used
To reduce by 3 kinds of mode computations, greatly reduce interframe encode complexity.Importantly, the algorithm does not need given threshold, this
Sample produces the probability of error with regard to very little, and when carrying out mode adjudging using textural characteristics, is made comparisons using optimal reference block,
So further ensure the accuracy of algorithm.
Brief description of the drawings
Fig. 1 is the AVS2 interframe fast mode judgment method streams based on Texture complication and time-space domain correlation of the present invention
Cheng Tu.
Embodiment
Below in conjunction with the accompanying drawings, the specific embodiment of the present invention is described further.
Method in the Inter-coded portions programming realization present invention of AVS2 identifying codes on a computer platform, specifically
Realize that step is as shown in Figure 1.Step is as follows:
(1) before being encoded to current CU, choose previous frame correspondence position same depth CU, and the same frame left side and
The CU of the same depth in top adjacent position as CU refer to, calculate it is each with reference to CU and current CU Texture complication, and from owning
Reference CU in selection one with current CU textures it is immediate be used as optimal reference block, record optimal reference block partition mode
For Mode0;
(2) rate distortion costs of NO_SPLIT patterns are calculated, and current CU optimum prediction mould BestMode is set to NO_
SPLIT patterns;
(3) the horizontal mean square error MSE of more current CUHorWith vertical mean square error MSEVerSize, if MSEHor<
MSEVer, jump to step (4);If MSEVer<MSEHor, jump to step (5);
(4) if Mode0 is one kind in three kinds of vertical division patterns VER_SYM, VER_LEFT, VER_RIGHT, calculate
Horizontal symmetrical pattern HOR_SYM and vertical symmetry pattern VER_SYM rate distortion costs, and compared with current best mode,
The pattern of Least-cost is selected as optimum prediction mode BestMode;Otherwise the rate of a calculated level symmetric pattern HOR_SYM
Distortion cost, and optimal mode BestMode is relatively obtained with current best mode;
(5) if Mode0 is one kind in three kinds of horizontal division patterns HOR_SYM, HOR_UP, HOR_DOWN, calculate and hang down
Straight symmetric pattern VER_SYM and horizontal symmetrical pattern HOR_SYM rate distortion costs, and compared with current best mode, choosing
The pattern of Least-cost is selected as optimum prediction mode BestMode;Otherwise the rate for only calculating vertical symmetry pattern VER_SYM is lost
True cost, and optimal mode BestMode is relatively obtained with current best mode;
(6) if optimal mode BestMode is NO_SPLIT patterns after step (4) or step (5), after terminating
Continuous asymmetric mode calculates, and jumps to step (9);
(7) if optimal mode BestMode is horizontal symmetrical pattern HOR_SYM after step (4) or step (5),
Calculated level asymmetric mode HOR_UP (HOR_DOWN) top position PU horizontal mean square error MSEUpUp(MSEDownUp) and under
Put PU horizontal mean square error MSE in orientationUpDown(MSEDownDown).If MSEUpUp<MSEUpDown(MSEDownDown<MSEDownUp), then
Mode computation is carried out to horizontal asymmetrical pattern HOR_UP (HOR_DOWN);Otherwise the pattern is skipped.
(8) if optimal mode BestMode is vertical symmetry pattern VER_SYM after step (4) or step (5),
Calculate vertical asymmetric mode VER_LEFT (VER_RIGHT) leftward positions PU vertical mean square error MSELeftLeft
(MSERightLeft) and right positions PU vertical mean square error MSELeftRight(MSERightRight).If MSELeftLeft<
MSELeftRight(MSERightRight<MSERightLeft), then row mode is entered to vertical asymmetric mode VER_LEFT (VER_RIGHT)
Calculate;Otherwise the pattern is skipped.
(9) more other pattern of surplus.
Multiple video sequences are encoded as experiment porch using AVS2 reference softwares RD14.0, and with original time
The algorithm for going through all patterns is compared, and calculates scramble time time, Y-PSNR PSNR under each quantization parameter QP
And coding bit rate Bit-rate change, shown in experimental result following table.
As can be known from the above table, set forth herein the inter-frame forecast mode based on time domain, spatial correlation and textural characteristics sentence
Certainly method, the interframe encode time can be significantly reduced.Average coding time is above 34% decline under different Q P, while Encoding
Can there was only minimum loss.
Claims (8)
1. a kind of quick decision method of AVS2 inter-frame forecast modes, it is characterised in that operating procedure is as follows:
1) before being encoded to current CU, the CU of previous frame correspondence position same depth, and the same frame left side and top are chosen
For the CU of the same depth in adjacent position as CU refer to, calculating is each with reference to CU and current CU Texture complication, and from all ginsengs
Examine and one and the immediate conduct optimal reference block of current CU textures are selected in CU, the partition mode of record optimal reference block is
Mode0;
2) rate distortion costs of NO_SPLIT patterns are calculated, and current CU optimum prediction mould BestMode is set to NO_SPLIT
Pattern;
3) the horizontal mean square error MSE of more current CUHorWith vertical mean square error MSEVerSize, if MSEHor<MSEVer, redirect
To step 4);If MSEVer<MSEHor, jump to step 5);
4) if Mode0 is one kind in three kinds of vertical division patterns VER_SYM, VER_LEFT, VER_RIGHT, calculated level
Symmetric pattern HOR_SYM and vertical symmetry pattern VER_SYM rate distortion costs, and compared with current best mode, selection
The pattern of Least-cost is as optimum prediction mode BestMode;Otherwise the rate distortion of a calculated level symmetric pattern HOR_SYM
Cost, and optimal mode BestMode is relatively obtained with current best mode;
5) if Mode0 is one kind in three kinds of horizontal division patterns HOR_SYM, HOR_UP, HOR_DOWN, vertical symmetry is calculated
Pattern VER_SYM and horizontal symmetrical pattern HOR_SYM rate distortion costs, and compared with current best mode, select cost
Minimum pattern is as optimum prediction mode BestMode;Otherwise vertical symmetry pattern VER_SYM rate distortion costs are only calculated,
And optimal mode BestMode is relatively obtained with current best mode;
If 6) optimal mode BestMode is NO_SPLIT patterns after step 4) or step 5), terminate follow-up asymmetric
Mode computation, jump to step 9);
If 7) optimal mode BestMode is horizontal symmetrical pattern HOR_SYM after step 4) or step 5), water is calculated
Flat asymmetric mode HOR_UP/HOR_DOWN top positions PU horizontal mean square error MSEUpUp/MSEDownUpWith lower position PU
Horizontal mean square error MSEUpDown/MSEDownDown;If MSEUpUp<MSEUpDown/MSEDownDown<MSEDownUp, then to horizontal non-right
Title pattern HOR_UP/HOR_DOWN carries out mode computation;Otherwise the pattern is skipped;
If 8) optimal mode BestMode is vertical symmetry pattern VER_SYM after step 4) or step 5), calculates and hang down
Straight asymmetric mode VER_LEFT/VER_RIGHT leftward positions PU vertical mean square error MSELeftLeft/MSERightLeftAnd the right side
Side position PU vertical mean square error MSELeftRight/MSERightRight;If MSELeftLeft<MSELeftRight/MSERightRight<
MSERightLeft, then mode computation is carried out to vertical asymmetric mode VER_LEFT/VER_RIGHT;Otherwise the pattern is skipped;
9) more other pattern of surplus.
A kind of 2. quick decision method of AVS2 inter-frame forecast modes according to claim 1, it is characterised in that the step
1) an optimal reference block is found for present encoding block from the adjacent block in time domain, spatial domain according to textural characteristics in, herein texture
Represented using least mean-square error MSE, be easy to calculate and the degree of accuracy is high;
Texture calculation formula is as follows:
<mrow>
<msub>
<mi>MSE</mi>
<mrow>
<mi>V</mi>
<mi>e</mi>
<mi>r</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mrow>
<mi>&alpha;</mi>
<mi>W</mi>
<mo>&times;</mo>
<mi>&beta;</mi>
<mi>H</mi>
</mrow>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>y</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mrow>
<mi>&alpha;</mi>
<mi>W</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>x</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mrow>
<mi>&beta;</mi>
<mi>H</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</munderover>
<msup>
<mrow>
<mo>(</mo>
<mi>p</mi>
<mo>(</mo>
<mrow>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
</mrow>
<mo>)</mo>
<mo>-</mo>
<msub>
<mi>m</mi>
<mi>y</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mrow>
<msub>
<mi>MSE</mi>
<mrow>
<mi>H</mi>
<mi>o</mi>
<mi>r</mi>
</mrow>
</msub>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mrow>
<mi>&alpha;</mi>
<mi>W</mi>
<mo>&times;</mo>
<mi>&beta;</mi>
<mi>H</mi>
</mrow>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>x</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mrow>
<mi>&beta;</mi>
<mi>H</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</munderover>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>y</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mrow>
<mi>&alpha;</mi>
<mi>W</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</munderover>
<msup>
<mrow>
<mo>(</mo>
<mi>p</mi>
<mo>(</mo>
<mrow>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
</mrow>
<mo>)</mo>
<mo>-</mo>
<msub>
<mi>m</mi>
<mi>x</mi>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
In formula, MSEVerAnd MSEHorRepresent the mean square errors of CU both vertically and horizontally;W and H is CU to be encoded respectively
Width and height, in units of pixel;α, β are coefficient values, for adjusting level and vertical direction under different demarcation pattern
On pixel number, calculate MSEVerAnd MSEHorWhen, α and β value are all 1;P [x] [y] represents the value of CU pixels;mxAnd my
Calculate MSEVerAnd MSEHorWhen represent respectively current CU (x+1)th row the average value of pixel and the pixel of y+1 row
Average value;
Abs=| MSEHor1-MSEHor2|+|MSEVer1-MSEVer2| 3)
And represent two encoding block texture degrees of closeness and use formula 3), the smaller textural characteristics for meaning that two encoding blocks of abs
It is closer.
A kind of 3. quick decision method of AVS2 inter-frame forecast modes according to claim 1, it is characterised in that:The step
2) NO_SPLIT patterns are set to optimal mode in, the calculation formula of its rate distortion costs is:
J(s,c,mode|QP,λmode)=SAD (s, c, mode | QP)+λmotionR(s,c,mode|QP) 4)
SAD represents that under conditions of given quantization parameter QP the predicting unit and the difference of reconstruction unit block of the pattern are put down in formula
Fang He;S is current prediction unit;C is reconstruction image block;Mode is selected interframe encoding mode;R is that institute is encoded under the pattern
The bit number needed;Lagrange factor λmotionIt is to be obtained according to quantization parameter QP.
A kind of 4. quick decision method of AVS2 inter-frame forecast modes according to claim 1, it is characterised in that:The step
3) determine to carry out horizontal symmetrical mode computation or vertical symmetry according to the textural characteristics of encoding block horizontally and vertically in
Mode computation, rather than two kinds of symmetric patterns are all calculated.
A kind of 5. quick decision method of AVS2 inter-frame forecast modes according to claim 1, it is characterised in that:The step
4) step 3) is avoided because error result caused by external factor as control with 5) the middle predictive mode by the use of optimal reference block
Subsequent process is had an impact.
A kind of 6. quick decision method of AVS2 inter-frame forecast modes according to claim 1, it is characterised in that:The step
If 6) rate distortion costs of horizontal symmetrical and vertical symmetry are all bigger than NO_SPLIT pattern in, horizontal asymmetrical and vertical non-
Symmetric pattern can not be used as optimal mode, therefore skip these patterns, avoid unnecessary calculating.
A kind of 7. quick decision method of AVS2 inter-frame forecast modes according to claim 1, it is characterised in that:The step
If 7) current best mode is horizontal symmetrical pattern in, predicting unit PU texture under further analysis level asymmetric mode
Complexity, and decided whether according to PU textural characteristics to carry out the calculating of each asymmetric mode, rather than all levels of traversal
Asymmetric mode;During the asymmetrical both of which HOR_UP and HOR_DOWN of analysis level, formula 2 is used), and use MSEUpUpWith
MSEUpDownRepresent the MSE of the block up and down under HOR_UP patterns respectively, α value is respectively 0.25 and 0.75 all for 1, β value;With
MSEDownUpAnd MSEDownDownThe MSE of the block up and down under HOR_DOWN patterns is represented respectively, and α value is all respectively for 1, β value
0.75 and 0.25;M when calculating this 4 PU MSExRepresent the average value of current PU (x+1)th row pixel.
A kind of 8. quick decision method of AVS2 inter-frame forecast modes according to claim 1, it is characterised in that:The step
If 8) current best mode is vertical symmetry pattern in, the texture of predicting unit PU under vertical asymmetric mode is further analyzed
Complexity, and decide whether according to PU textural characteristics to carry out the calculating of each asymmetric mode, rather than traversal is all vertical
Asymmetric mode;When analyzing vertical asymmetrical both of which VER_LEFT and VER_RIGHT, formula 1 is used), it is used in combination
MSELeftLeftAnd MSELeftRightRepresenting the MSE of the left and right block under VER_LEFT patterns respectively, α value is respectively 0.25 and 0.75,
β value is all 1;Use MSERightLeftAnd MSERightRightThe MSE of the left and right block under VER_RIGHT patterns, α value are represented respectively
Respectively 0.75 and 0.25, β value all be 1;M when calculating this 4 PU MSEyRepresent being averaged for current PU y+1 row pixels
Value.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710315131.2A CN107343198A (en) | 2017-05-08 | 2017-05-08 | A kind of quick decision method of AVS2 inter-frame forecast modes |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710315131.2A CN107343198A (en) | 2017-05-08 | 2017-05-08 | A kind of quick decision method of AVS2 inter-frame forecast modes |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107343198A true CN107343198A (en) | 2017-11-10 |
Family
ID=60220207
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710315131.2A Pending CN107343198A (en) | 2017-05-08 | 2017-05-08 | A kind of quick decision method of AVS2 inter-frame forecast modes |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107343198A (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108124154A (en) * | 2017-12-28 | 2018-06-05 | 北京数码视讯科技股份有限公司 | Fast selecting method, device and the electronic equipment of inter-frame forecast mode |
CN109151467A (en) * | 2018-09-10 | 2019-01-04 | 重庆邮电大学 | Screen content based on image block activity encodes interframe mode quick selecting method |
CN109547798A (en) * | 2018-12-17 | 2019-03-29 | 杭州当虹科技股份有限公司 | A kind of quick HEVC inter-frame mode selecting method |
CN111669602A (en) * | 2020-06-04 | 2020-09-15 | 北京大学深圳研究生院 | Method and device for dividing coding unit, coder and storage medium |
CN111683245A (en) * | 2020-06-23 | 2020-09-18 | 北京工业职业技术学院 | Texture similarity based CU partition decision |
CN112804524A (en) * | 2019-11-13 | 2021-05-14 | 北京大学 | Intra-frame fast mode decision method for AVS2 |
CN112866691A (en) * | 2019-11-27 | 2021-05-28 | 北京大学 | Inter-frame fast mode decision method for video coding |
WO2021164323A1 (en) * | 2020-02-17 | 2021-08-26 | 腾讯科技(深圳)有限公司 | Coding mode selection method and apparatus, and electronic device and computer readable medium |
CN113542753A (en) * | 2021-07-27 | 2021-10-22 | 杭州当虹科技股份有限公司 | AVS3 video coding method and encoder |
Citations (3)
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 |
CN103384325A (en) * | 2013-02-22 | 2013-11-06 | 张新安 | Quick inter-frame prediction mode selection method for AVS-M video coding |
CN105141954A (en) * | 2015-08-19 | 2015-12-09 | 浙江工业大学 | HEVC interframe coding quick mode selection method |
-
2017
- 2017-05-08 CN CN201710315131.2A patent/CN107343198A/en active Pending
Patent Citations (3)
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 |
CN103384325A (en) * | 2013-02-22 | 2013-11-06 | 张新安 | Quick inter-frame prediction mode selection method for AVS-M video coding |
CN105141954A (en) * | 2015-08-19 | 2015-12-09 | 浙江工业大学 | HEVC interframe coding quick mode selection method |
Non-Patent Citations (1)
Title |
---|
产小红,滕国伟,王国中,赵海武,李国平: "AVS2帧间预测模式快速判决方法", 《电视技术》 * |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108124154A (en) * | 2017-12-28 | 2018-06-05 | 北京数码视讯科技股份有限公司 | Fast selecting method, device and the electronic equipment of inter-frame forecast mode |
CN109151467A (en) * | 2018-09-10 | 2019-01-04 | 重庆邮电大学 | Screen content based on image block activity encodes interframe mode quick selecting method |
CN109547798A (en) * | 2018-12-17 | 2019-03-29 | 杭州当虹科技股份有限公司 | A kind of quick HEVC inter-frame mode selecting method |
CN109547798B (en) * | 2018-12-17 | 2020-09-18 | 杭州当虹科技股份有限公司 | Rapid HEVC inter-frame mode selection method |
CN112804524B (en) * | 2019-11-13 | 2022-11-25 | 北京大学 | Intra-frame fast mode decision method for AVS2 |
CN112804524A (en) * | 2019-11-13 | 2021-05-14 | 北京大学 | Intra-frame fast mode decision method for AVS2 |
CN112866691B (en) * | 2019-11-27 | 2022-09-27 | 北京大学 | Inter-frame fast mode decision method for video coding |
CN112866691A (en) * | 2019-11-27 | 2021-05-28 | 北京大学 | Inter-frame fast mode decision method for video coding |
WO2021164323A1 (en) * | 2020-02-17 | 2021-08-26 | 腾讯科技(深圳)有限公司 | Coding mode selection method and apparatus, and electronic device and computer readable medium |
CN111669602A (en) * | 2020-06-04 | 2020-09-15 | 北京大学深圳研究生院 | Method and device for dividing coding unit, coder and storage medium |
CN111683245B (en) * | 2020-06-23 | 2022-07-22 | 北京工业职业技术学院 | Texture similarity based CU partition decision |
CN111683245A (en) * | 2020-06-23 | 2020-09-18 | 北京工业职业技术学院 | Texture similarity based CU partition decision |
CN113542753B (en) * | 2021-07-27 | 2022-07-08 | 杭州当虹科技股份有限公司 | AVS3 video coding method and encoder |
CN113542753A (en) * | 2021-07-27 | 2021-10-22 | 杭州当虹科技股份有限公司 | AVS3 video coding method and encoder |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107343198A (en) | A kind of quick decision method of AVS2 inter-frame forecast modes | |
CN110087087B (en) | VVC inter-frame coding unit prediction mode early decision and block division early termination method | |
CN100534194C (en) | Methods and apparatus for video intraprediction encoding and decoding | |
CN103517069B (en) | A kind of HEVC intra-frame prediction quick mode selection method based on texture analysis | |
WO2018010492A1 (en) | Rapid decision making method for intra-frame prediction mode in video coding | |
CN101964906B (en) | Rapid intra-frame prediction method and device based on texture characteristics | |
CN102158709B (en) | Derivable motion compensation prediction method of decoding terminal | |
CN101309421B (en) | Intra-frame prediction mode selection method | |
CN106961606A (en) | The HEVC intra-frame encoding mode systems of selection of feature are divided based on texture | |
CN105208387B (en) | A kind of HEVC Adaptive Mode Selection Method for Intra-Prediction | |
CN100534195C (en) | Fast inter-frame mode adjudging method capable of fusing multi-reference frame selection and motion estimation | |
CN105141954A (en) | HEVC interframe coding quick mode selection method | |
CN105681808B (en) | A kind of high-speed decision method of SCC interframe encodes unit mode | |
CN105187826B (en) | For the fast intra mode decision method of high efficiency video encoding standard | |
CN105898332B (en) | For the fast deep figure frame mode decision method of 3D-HEVC coding standards | |
CN101304529A (en) | Method and device for selecting macro block pattern | |
CN103997645A (en) | Quick HEVC intra-frame coding unit and pattern decision method | |
CN100596202C (en) | Fast mode selection method in frame | |
CN106791828A (en) | High performance video code-transferring method and its transcoder based on machine learning | |
CN102186081A (en) | H.264 intra-frame mode selection method based on gradient vector | |
CN102164283A (en) | Sub-pixel motion estimation method based on AVS (audio video standard) | |
Huang et al. | Rate-distortion-complexity optimized coding mode decision for HEVC | |
JP5566786B2 (en) | Error absolute value sum estimation system and estimation method | |
CN101883275B (en) | Video coding method | |
CN101867818B (en) | Selection method and device of macroblock mode |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171110 |