CN102387364A - Fast intra-frame mode selecting algorithm - Google Patents

Fast intra-frame mode selecting algorithm Download PDF

Info

Publication number
CN102387364A
CN102387364A CN2011103318046A CN201110331804A CN102387364A CN 102387364 A CN102387364 A CN 102387364A CN 2011103318046 A CN2011103318046 A CN 2011103318046A CN 201110331804 A CN201110331804 A CN 201110331804A CN 102387364 A CN102387364 A CN 102387364A
Authority
CN
China
Prior art keywords
pattern
luminance block
prediction mode
optimum prediction
block
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.)
Granted
Application number
CN2011103318046A
Other languages
Chinese (zh)
Other versions
CN102387364B (en
Inventor
宋雪桦
包祥
顾金
谢桂莹
王昌达
吴问云
路敏
吴合生
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jiangsu University
Original Assignee
Jiangsu University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Jiangsu University filed Critical Jiangsu University
Priority to CN 201110331804 priority Critical patent/CN102387364B/en
Publication of CN102387364A publication Critical patent/CN102387364A/en
Application granted granted Critical
Publication of CN102387364B publication Critical patent/CN102387364B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Landscapes

  • Compression Or Coding Systems Of Tv Signals (AREA)

Abstract

The invention discloses a fast intra-frame mode selecting algorithm. The fast intra-frame mode selecting algorithm firstly carries out judgment in horizontal and vertical directions according to the judgment criteria of a block in the horizontal and vertical directions so as to fast determine whether a luminance block belongs to a mode 0 or a mode 1, judges the mode of the luminance block by using the space relevance of adjacent blocks, and finally fast selects an algorithm to carry out the judgment by using edge direction histograms; and the algorithm is suitable for carrying out fast calculation of computers and is lower in complexity. The algorithm comprehensively utilizes the relevance of the modes of the adjacent blocks and the directivity of luminance changing in the blocks, thus greatly shortening the calculated quantity and having good practicality.

Description

A kind of fast intra mode selection algorithm
Technical field
The invention belongs to the audio/video encoding/decoding technical field, according to own characteristic and combine infra-prediction techniques and H.264/AVC some intraframe prediction algorithms of AVS, thus the fast intra mode selection algorithm of a kind of low complex degree that proposes.
Background technology
AVS (Audio Video coding Standard) is that China independently formulates, and has the audio/video coding technical standard of independent intellectual property right.AVS has adopted a series of advanced technologies to improve video compression efficiency; Comprise: infra-frame prediction, inter prediction, estimation, motion compensation, change quantization, entropy coding, loop filtering, rate-distortion optimization (Rate Distortion Optimization, key technology such as RDO).Wherein the amount of calculation of RDO is very big, and various coding mode all will carry out RDO computing consuming time in the hope of optimal mode.Frame mode prediction full-search algorithm has guaranteed best compression efficiency and picture quality; But huge amount of calculation becomes its fatal drawback; Add interframe encode and intraframe coding and all will carry out infra-frame prediction, infra-frame prediction also is module quite consuming time in the compression encoding process.Therefore, the complexity that how to reduce the frame mode selection algorithm is to need a major issue considering.
At present, both at home and abroad frame mode selection algorithm has H.264/AVC been carried out a large amount of research, and obtained significant achievement.Wherein more typical algorithm has: the fast intra mode selection algorithm that proposes of people such as Feng Pan based on local edge direction information; Utilize the Sobel operator to calculate the edge direction vector of each pixel the earliest; Utilize the edge direction vector to try to achieve the edge orientation histogram of each piece then; Through this histogrammic characteristic distributions, select the bigger several patterns of possibility at last as candidate pattern. utilizes the correlation between the rate distortion costs under the different directions predictive mode; Selectively skip unlikely predictive mode, thereby dwindle the candidate pattern scope.The fast frame inner estimation mode that combines the algorithm of Feng Pan to propose is selected new method; It has utilized the characteristic of the reference pixel of block edge; Absolute difference mean value is set a threshold value, confirm optimum prediction mode in advance.These algorithm common problems are exactly that algorithm complex is higher, and operand is big, are unfavorable in the hardware platform real-time implementation.
The present invention is through the infra-prediction techniques of research AVS and some intraframe prediction algorithms H.264/AVC; And a kind of fast intra mode selection algorithm of low complex degree proposed according to own characteristic; Make full use of the correlation of adjacent block pattern and the directivity that the interior brightness of piece changes, shortened amount of calculation greatly.
Summary of the invention
The purpose of this invention is to provide a kind of correlation of adjacent block optimal mode and directivity of the interior brightness variation of piece utilized and carry out the pattern anticipation; And combination Feng Pan algorithm; Dwindle the candidate pattern scope; Under the prerequisite that does not influence coding efficiency, both do not increased complicated calculations, also effectively improved coding rate.
1. technical scheme of the present invention is: a kind of fast intra mode selection algorithm is characterized in that may further comprise the steps:
Step 1 adopts piece horizontal direction and vertical direction judgment criterion, judges whether the optimum prediction mode of luminance block belongs to pattern 0 or pattern 1, otherwise gets into step 2;
Step 2 utilizes the spatial coherence of adjacent block to judge whether luminance block belongs to pattern 2, mode 3 or pattern 4;
If the left side piece of step 2.1 luminance block and top piece do not exist, judge that the optimum prediction mode of luminance block is a pattern 2, otherwise execution in step 2.2;
If step 2.2 has only the top piece of luminance block to exist, the candidate pattern of then selecting luminance block is the pattern or the DC pattern of said top piece; Calculate the matching error value J of the pattern and pattern 2 correspondences of said top piece OnAnd J DC, select said J OnAnd J DCThe corresponding pattern of smaller value be the optimum prediction mode of luminance block; Otherwise execution in step 2.3;
If step 2.3 has only the left side piece of luminance block to exist, the candidate pattern of luminance block is the pattern or the DC pattern of said left side piece; Calculate the matching error value J of the pattern and pattern 2 correspondences of said left side piece A left sideAnd J DC, select said J A left sideAnd J DCThe corresponding pattern of smaller value be the optimum prediction mode of luminance block; Otherwise execution in step 2.4;
If the left side piece of step 2.4 luminance block and top piece all exist, selecting the pattern of left side piece or the pattern of top piece is the candidate pattern of luminance block, calculated threshold
Figure 927955DEST_PATH_IMAGE004
, get J A left sideAnd J OnSmaller value with
Figure 2011103318046100002DEST_PATH_IMAGE005
Compare, if less than
Figure 271474DEST_PATH_IMAGE005
, the pattern that then said smaller value is corresponding is the optimum prediction mode of luminance block, otherwise execution in step 3.
Step 3 adopts the optimum prediction mode of judging luminance block based on the fast selection algorithm of edge orientation histogram;
It is that 2:1 carries out sub-sampling that the original image pixels of step 3.1 pair input luminance block is pressed the sampling ratio;
Step 3.2 definition edge vectors makes up edge orientation histogram;
Step 3.3 is through judging whether luma block edge direction histogram has unimodality to confirm the optimum prediction mode of luminance block;
Piece horizontal direction and vertical direction judgment criterion in the said step 1 are: the difference that respectively 16 pixels are right on calculated level direction and the vertical direction and, and the jointing edge orientation angle judges that its concrete steps are:
Step 1.1 is divided into 4 4x4 pieces with the 8x8 luminance block; Be labeled as
Figure 666683DEST_PATH_IMAGE006
,
Figure 2011103318046100002DEST_PATH_IMAGE007
,
Figure 605689DEST_PATH_IMAGE008
,
Figure 2011103318046100002DEST_PATH_IMAGE009
respectively, it is defined as:
Figure 128286DEST_PATH_IMAGE010
Figure 2011103318046100002DEST_PATH_IMAGE011
Figure 333003DEST_PATH_IMAGE012
Figure 2011103318046100002DEST_PATH_IMAGE013
S wherein X, yBe the right value of each pixel;
Step 1.2 is set the vertical direction parameter and the horizontal direction parameter is respectively and
Figure 2011103318046100002DEST_PATH_IMAGE015
, and it is defined as:
Figure 13831DEST_PATH_IMAGE016
Figure 2011103318046100002DEST_PATH_IMAGE017
Step 1.3 utilizes the edge direction angle information to judge edge direction; Make
Figure 131828DEST_PATH_IMAGE018
,
Figure 2011103318046100002DEST_PATH_IMAGE019
represent vertical edge orientation angle and horizontal edge orientation angle respectively, be defined as:
Figure 512256DEST_PATH_IMAGE020
Figure 2011103318046100002DEST_PATH_IMAGE021
Step 1.4 setting threshold
Figure 186951DEST_PATH_IMAGE022
, judge according to rule:
As
Figure 2011103318046100002DEST_PATH_IMAGE023
; And when
Figure 897287DEST_PATH_IMAGE024
; Be horizontal direction, judge that the optimum prediction mode of luminance block is a pattern 0;
As ; And when
Figure 953230DEST_PATH_IMAGE026
; Be vertical direction, judge that the optimum prediction mode of luminance block is a pattern 1.
The concrete steps of said step 3.3 are:
Amplitude maximum in the step 3.3.1 edge calculation direction histogram
Figure 2011103318046100002DEST_PATH_IMAGE027
, its corresponding pattern are
Figure 132539DEST_PATH_IMAGE028
;
If step 3.3.2 works as
Figure 594613DEST_PATH_IMAGE027
less than threshold value
Figure DEST_PATH_IMAGE029
; Selecting its optimal mode is pattern 2, otherwise execution in step 3.3.3;
Amplitude in the step 3.3.3 compute histograms time maximum
Figure 864183DEST_PATH_IMAGE030
, its corresponding pattern is
Figure DEST_PATH_IMAGE031
;
Step 3.3.4 calculates poor
Figure 137032DEST_PATH_IMAGE032
of the amplitude of two kinds of patterns; During greater than threshold value
Figure 52904DEST_PATH_IMAGE034
, selecting the corresponding pattern
Figure 69402DEST_PATH_IMAGE028
of amplitude maximum is the optimum prediction mode of luminance block as
Figure DEST_PATH_IMAGE033
; Otherwise comparison pattern
Figure 255795DEST_PATH_IMAGE028
and pattern 2 corresponding matching error values, selecting the corresponding pattern of less matching error value is the optimum prediction mode of luminance block.
The invention has the beneficial effects as follows: at first carry out the judgement of level and vertical direction according to piece horizontal direction and vertical direction judgment criterion; Because the likelihood ratio mode 3 and pattern 4 height of pattern 0 and pattern 1 appearance; And pattern 0 has very strong directivity with pattern 1; Judgement is got up also relatively easy, can utilize this step to confirm that fast luminance block belongs to pattern 0 or pattern 1, has reduced the complexity of algorithm; Moreover the spatial coherence that utilizes adjacent block judges the pattern of luminance block, and this algorithm is suitable to carry out the quick calculating of computer, and its algorithm complexity is also lower; The fast selection algorithm that is used in edge orientation histogram is at last judged.It has fully utilized the correlation of adjacent block pattern and the directivity that the interior brightness of piece changes, and has shortened amount of calculation greatly, and has had good practicability.
Description of drawings
Fig. 1 fast intra mode selection algorithm flow chart;
Fig. 2 is based on edge orientation histogram fast selection algorithm flow chart.
Embodiment
Further specify below in conjunction with accompanying drawing and instance.
5 kinds of predictive modes of AVS luminance block; Be respectively pattern 0 (vertically), pattern 1 (level), pattern 2 (DC), mode 3 (left side down), pattern 4 (bottom right); Wherein pattern 0, pattern 1, mode 3, pattern 4 these four kinds of patterns have directivity; Can come Fast estimation intra-block edge direction according to the directivity that brightness in the piece changes, so can judge to above-mentioned four directive patterns.Generally speaking, likelihood ratio mode 3 and pattern 4 height that pattern 0 and pattern 1 occur, and pattern 0 has very strong directivity with pattern 1 are so that judgement is got up is also relatively easy.
Carry out the judgement of level and vertical direction according to piece horizontal direction and vertical direction judgment criterion, can confirm whether the luminance block that will judge belongs to this two kinds of predictive modes.
A, differentiate luminance block and whether belong to pattern 0 or pattern 1: the difference that 16 pixels are right on calculated level direction and the vertical direction respectively and, and the jointing edge orientation angle is judged its direction.
A 8x8 luminance block is divided into 4 4x4 pieces;
Figure 965125DEST_PATH_IMAGE006
,
Figure 915763DEST_PATH_IMAGE007
,
Figure 985219DEST_PATH_IMAGE008
, are made in order respectively, define as follows:
Figure 217935DEST_PATH_IMAGE010
Figure 702585DEST_PATH_IMAGE012
Figure 175155DEST_PATH_IMAGE013
S wherein X, yBe the right value of each pixel;
Pattern 0 has very strong directivity with pattern 1; In order to obtain the edge directional information of this luminance block; Make
Figure 475555DEST_PATH_IMAGE014
;
Figure 338469DEST_PATH_IMAGE015
representes vertical direction parameter and horizontal direction parameter respectively, defines as follows:
Figure 312689DEST_PATH_IMAGE017
Utilize the edge direction angle information can judge edge direction more accurately; Make
Figure 534723DEST_PATH_IMAGE018
,
Figure 134200DEST_PATH_IMAGE019
represent vertical edge orientation angle and horizontal edge orientation angle respectively, definition as follows:
Figure 162199DEST_PATH_IMAGE020
Figure 281465DEST_PATH_IMAGE021
Wherein
Figure 147835DEST_PATH_IMAGE018
,
Figure 985341DEST_PATH_IMAGE019
middle the greater are as denominator; Setting threshold
Figure 551451DEST_PATH_IMAGE022
, carry out following rule and judge according to
Figure 40070DEST_PATH_IMAGE014
,
Figure 338328DEST_PATH_IMAGE015
,
Figure 725447DEST_PATH_IMAGE018
,
Figure 783664DEST_PATH_IMAGE019
then:
As ; When
Figure 408997DEST_PATH_IMAGE024
; Being horizontal direction, is pattern 0 for the luminance block optimum prediction mode then;
As ; When
Figure 315959DEST_PATH_IMAGE026
; Be vertical direction, then spending the piece optimum prediction mode is pattern 1.
B, utilize the spatial coherence of adjacent block to judge whether luminance block belongs to pattern 2, mode 3 or pattern 4.
Judgment criterion is following:
(1) if the left side piece of luminance block and top piece do not exist, establish this moment for state one, think that then the optimum prediction mode of luminance block is pattern 2 (DC), finish search, otherwise execution in step (2);
(2) if having only the top piece of luminance block to exist (because exist; So the coding of top piece and predictive mode all are known; Left side block prediction mode also is the same), establishing this moment is state two, then the candidate pattern of luminance block is top block mode or DC pattern; Calculate the corresponding matching error value J of top block mode and DC pattern OnAnd J DC, select J OnAnd J DCThe corresponding pattern of smaller value be the optimum prediction mode of luminance block; Otherwise execution in step (3);
(3) if the left side piece that has only of luminance block exists, establishing this moment is state three, and the candidate pattern of then selecting luminance block is left side block mode or DC pattern; Calculate the corresponding matching error value J of left side block mode and DC pattern A left sideAnd J DC, select J A left sideAnd J DCThe corresponding pattern of smaller value be the optimum prediction mode of luminance block; Otherwise execution in step (4);
(4) if the left side piece of luminance block and top piece all exist, establishing this moment is state four, computation of match errors value J value,
Figure DEST_PATH_IMAGE035
Wherein: the quadratic sum of the respective pixel value difference value of original block and reconstructed blocks under
Figure 15056DEST_PATH_IMAGE036
expression associative mode; The code check of the code stream that obtains behind
Figure DEST_PATH_IMAGE037
expression residual block process entropy coding;
Figure 45329DEST_PATH_IMAGE038
takes advantage of the factor for Lagrange, and be relevant with quantization parameter.
Figure 829876DEST_PATH_IMAGE036
defines as follows:
Figure DEST_PATH_IMAGE039
Wherein: the pixel value of
Figure 479163DEST_PATH_IMAGE040
expression reconstructed blocks, the pixel value of
Figure DEST_PATH_IMAGE041
expression original block;
According to J A left sideAnd J OnCalculated threshold
Figure 468985DEST_PATH_IMAGE005
, and with left side block mode or the top block mode candidate pattern as luminance block.
Figure 296258DEST_PATH_IMAGE042
J wherein A left sideAnd J OnThe matching error value of representing left side piece and top piece respectively.Smaller in the two and
Figure 879686DEST_PATH_IMAGE005
are compared; If less than
Figure 67085DEST_PATH_IMAGE005
; Then selecting the pattern of this smaller value institute corresponding blocks is the luminance block optimum prediction mode; Finish search, otherwise adopt the fast selection algorithm among the C based on edge orientation histogram.
C. judge the optimum prediction mode of luminance block based on the fast selection algorithm of edge orientation histogram;
The first step: the pixel to luminance block is sampled.
The original image pixels point of luminance block to input carries out the 2:1 sub-sampling, makes that pixel number after the sampling is half the before the sampling.Because the pixel value behind the sub-sampling is to be asked on average by the value of adjacent two pixels in the left and right sides to obtain; And the spatial coherence of neighbor pixel is very strong; So the data after sub-sampling is handled have not only kept the data characteristics of original image, also reduced computing time significantly simultaneously.
Second step: definition edge direction vector, structure edge orientation histogram.
The present invention utilizes the sobel operator to calculate the edge direction vector of each pixel.The sobel operator has two convolution kernels to represent the difference degree of horizontal direction and vertical direction respectively.For any pixel after the luminance block sampling; Defining its edge vectors is:
Figure 239309DEST_PATH_IMAGE044
, wherein:
Figure DEST_PATH_IMAGE045
Figure 175166DEST_PATH_IMAGE046
Figure DEST_PATH_IMAGE047
and representes vertical direction and horizontal direction component respectively, and the amplitude of edge vectors can be similar to and be defined as:
Figure DEST_PATH_IMAGE049
The edge vectors direction can be defined as:
Figure 361614DEST_PATH_IMAGE050
，?
Figure DEST_PATH_IMAGE051
Because pattern 0 (vertically), pattern 1 (level), mode 3 (left side down) and pattern 4 (bottom right) are based on the predictive mode of direction in the intra prediction mode of luminance block; Select interval
Figure 827493DEST_PATH_IMAGE052
so be divided into four infra-frame predictions to the whole prediction direction; Corresponding interval interior pixel set expression with ; The forecast interval
Figure 495103DEST_PATH_IMAGE054
that for example comprises horizontal forecast pattern 1; The pixel set that then belongs to this pattern is for
Figure DEST_PATH_IMAGE055
, and other by that analogy.
Make
Figure 725227DEST_PATH_IMAGE056
here; Add and be worth for every kind of predictive mode in the histogram is pairing; Promptly for the pixel after the sampling; Judge respectively which kind of pattern it belongs to, the modulo addition with the vector of model identical obtains then.
Figure 208424DEST_PATH_IMAGE058
Wherein
Figure 89661DEST_PATH_IMAGE053
is the set of the pixel of certain pattern, sets up edge orientation histogram thus.
The 3rd step: the optimum prediction mode of judging and select luminance block.
Owing to all be to select a maximum pattern of amplitude in the edge orientation histogram in the Pan scheme; Add a DC pattern; But actual DC pattern is only effective relatively when the directivity of piece is unconspicuous, so thereby can whether obviously confirm to want to adopt the DC pattern through the directivity of judging this luminance block.Generally, front-to-back ratio significantly piece can show apparent in view unimodality on histogram, therefore, judge whether a luminance block directivity obviously judges with regard to being converted into whether its edge orientation histogram has the problem of unimodality.The present invention judges through following method:
(1) obtain amplitude maximum in the edge orientation histogram, corresponding pattern be
Figure 741222DEST_PATH_IMAGE028
(
Figure 724222DEST_PATH_IMAGE028
be in the pattern 0-pattern 4 a kind of)
(2) as
Figure 819565DEST_PATH_IMAGE027
during less than threshold value (
Figure 394083DEST_PATH_IMAGE029
gets 720); Direct employing pattern 2, otherwise carry out next step;
(3) ask amplitude time maximum
Figure 848067DEST_PATH_IMAGE030
in the histogram, its corresponding pattern be (
Figure 14048DEST_PATH_IMAGE031
be in the pattern 0-pattern 4 a kind of); Then
Figure 459942DEST_PATH_IMAGE032
Figure 214271DEST_PATH_IMAGE033
is amplitude poor of two kinds of patterns; As
Figure 888966DEST_PATH_IMAGE033
during greater than threshold value
Figure 100767DEST_PATH_IMAGE034
(
Figure 530611DEST_PATH_IMAGE034
gets 950); Thinking has unimodality, directly selects the corresponding pattern
Figure 444341DEST_PATH_IMAGE028
of amplitude maximum to be the optimum prediction mode of luminance block; Otherwise compare
Figure 171994DEST_PATH_IMAGE028
pattern and pattern 2 corresponding matching error values, selecting the corresponding pattern of wherein less matching error value is the optimum prediction mode of luminance block.
is to be got through a large amount of experiment tests; This patent provides the reference value of each threshold value;
Figure 505652DEST_PATH_IMAGE029
gets 720, and
Figure 522149DEST_PATH_IMAGE034
gets 950.
Algorithm of the present invention is adapted to the Rapid Realization of computer, and its algorithm steps is as shown in Figure 1:
Step S101 judges whether the change direction of current luminance block is horizontal direction or vertical direction, if, execution in step S102 then, otherwise execution in step S103;
Then to declare optimum prediction mode be pattern 0 to step S102 if the change direction of luminance block is a horizontal direction, and then to declare optimum prediction mode be pattern 1 if its change direction is a vertical direction, finishes search;
Step S103 judges whether the left adjacent block of current luminance block and last adjacent block all exist, if not, execution in step S104 then, otherwise execution in step S105;
Step S104 confirms the optimal mode that each state is corresponding and finishes search;
If the left adjacent block of the current luminance block of step S105 and last adjacent block all exist, matching error value that smaller in left side piece and the top block mode is corresponding and threshold value TH2 are relatively;
If step S106 is less than threshold value
Figure DEST_PATH_IMAGE059
; Execution in step S107 then, otherwise execution in step S108;
It is the optimum prediction mode and the end search of luminance block that step S107 selects the corresponding pattern of less matching error value;
Step S108 adopts and carries out model selection based on the edge orientation histogram fast selection algorithm.
Specifically as shown in Figure 2 based on the edge orientation histogram fast selection algorithm:
Step S201 carries out the 2:1 sub-sampling to the original image pixels point of luminance block;
Step S202 chooses the pixel behind first sampled point;
Does step S203 judge whether to travel through all sub-sampling pixels of this luminance block? (judging) through whether there being next pixel among the step S202 if, execution in step S207 then, otherwise, execution in step S204;
The edge direction vector of step S204 calculating pixel point;
Step S205 calculates and makes up edge orientation histogram;
Step S206 continues to choose next pixel, repeating step 203;
Step S207 chooses the corresponding pattern
Figure 650828DEST_PATH_IMAGE060
of amplitude maximum in the edge orientation histogram, and
Edge direction histogram amplitude times the maximum
Figure 555461DEST_PATH_IMAGE030
corresponding mode
Figure 437967DEST_PATH_IMAGE031
;
Whether step S208 judges
Figure 728134DEST_PATH_IMAGE027
less than threshold value
Figure 857633DEST_PATH_IMAGE029
; If; Execution in step S209 then, otherwise execution in step S210;
Step S209 confirms that the optimum prediction mode of luminance block is a pattern 2, finishes search;
Whether step S210 judges
Figure 295567DEST_PATH_IMAGE032
greater than threshold value ; If; Execution in step S211 then, otherwise execution in step S212;
The optimum prediction mode that step S211 confirms luminance block is , finishes search;
Middle corresponding matching error value J smaller is the optimum prediction mode of luminance block to step S212 preference pattern 2 with
Figure 662723DEST_PATH_IMAGE060
, finishes search.

Claims (5)

1. fast intra mode selection algorithm is characterized in that may further comprise the steps:
Step 1 adopts piece horizontal direction and vertical direction judgment criterion, judges whether the optimum prediction mode of luminance block belongs to pattern 0 or pattern 1, otherwise gets into step 2;
Step 2 utilizes the spatial coherence of adjacent block to judge whether luminance block belongs to pattern 2, mode 3 or pattern 4;
If the left side piece of step 2.1 luminance block and top piece do not exist, judge that the optimum prediction mode of luminance block is a pattern 2, otherwise execution in step 2.2;
If step 2.2 has only the top piece of luminance block to exist, the candidate pattern of then selecting luminance block is the pattern or the DC pattern of said top piece; Calculate the matching error value J of the pattern and pattern 2 correspondences of said top piece OnAnd J DC, select said J OnAnd J DCThe corresponding pattern of smaller value be the optimum prediction mode of luminance block; Otherwise execution in step 2.3;
If step 2.3 has only the left side piece of luminance block to exist, the candidate pattern of luminance block is the pattern or the DC pattern of said left side piece; Calculate the matching error value J of the pattern and pattern 2 correspondences of said left side piece A left sideAnd J DC, select said J A left sideAnd J DCThe corresponding pattern of smaller value be the optimum prediction mode of luminance block; Otherwise execution in step 2.4;
If the left side piece of step 2.4 luminance block and top piece all exist, selecting the pattern of left side piece or the pattern of top piece is the candidate pattern of luminance block, calculated threshold
Figure 2011103318046100001DEST_PATH_IMAGE002
, get J A left sideAnd J OnSmaller value with
Figure 2011103318046100001DEST_PATH_IMAGE004
Compare, if less than
Figure 30438DEST_PATH_IMAGE004
, the pattern that then said smaller value is corresponding is the optimum prediction mode of luminance block, otherwise execution in step 3.
2. step 3 adopts the optimum prediction mode of judging luminance block based on the fast selection algorithm of edge orientation histogram;
The original image pixels point of step 3.1 pair input luminance block carries out sub-sampling;
Step 3.2 definition edge vectors makes up edge orientation histogram;
Step 3.3 is through judging whether luma block edge direction histogram has unimodality to confirm the optimum prediction mode of luminance block;
A kind of fast intra mode selection algorithm according to claim 1; It is characterized in that: piece horizontal direction and vertical direction judgment criterion in the said step 1 are: the difference that respectively 16 pixels are right on calculated level direction and the vertical direction and; And the jointing edge orientation angle judges that its concrete steps are:
Step 1.1 is divided into 4 4x4 pieces with the 8x8 luminance block; Be labeled as
Figure 2011103318046100001DEST_PATH_IMAGE006
,
Figure 2011103318046100001DEST_PATH_IMAGE008
, ,
Figure 2011103318046100001DEST_PATH_IMAGE012
respectively, it is defined as:
Figure 2011103318046100001DEST_PATH_IMAGE014
Figure 2011103318046100001DEST_PATH_IMAGE016
Figure 2011103318046100001DEST_PATH_IMAGE018
Figure 2011103318046100001DEST_PATH_IMAGE020
S wherein X, yBe the right value of each pixel;
Step 1.2 is set the vertical direction parameter and the horizontal direction parameter is respectively
Figure 2011103318046100001DEST_PATH_IMAGE022
and , and it is defined as:
Figure 2011103318046100001DEST_PATH_IMAGE026
Figure 2011103318046100001DEST_PATH_IMAGE028
Step 1.3 utilizes the edge direction angle information to judge edge direction; Make
Figure 2011103318046100001DEST_PATH_IMAGE030
, represent vertical edge orientation angle and horizontal edge orientation angle respectively, be defined as:
Figure 2011103318046100001DEST_PATH_IMAGE034
Figure 2011103318046100001DEST_PATH_IMAGE036
Step 1.4 setting threshold
Figure 2011103318046100001DEST_PATH_IMAGE038
, judge according to rule:
As
Figure 2011103318046100001DEST_PATH_IMAGE040
; And when
Figure 2011103318046100001DEST_PATH_IMAGE042
; Be horizontal direction, judge that the optimum prediction mode of luminance block is a pattern 0;
As
Figure 2011103318046100001DEST_PATH_IMAGE044
; And when
Figure 2011103318046100001DEST_PATH_IMAGE046
; Be vertical direction, judge that the optimum prediction mode of luminance block is a pattern 1.
3. a kind of fast intra mode selection algorithm according to claim 1 is characterized in that: the sampling ratio of said step 3.1 is 2:1.
4. a kind of fast intra mode selection algorithm according to claim 1, it is characterized in that: the concrete steps of said step 3.3 are:
Amplitude maximum in the step 3.3.1 edge calculation direction histogram
Figure 2011103318046100001DEST_PATH_IMAGE048
, its corresponding pattern are
Figure 2011103318046100001DEST_PATH_IMAGE050
;
If step 3.3.2 works as
Figure 383796DEST_PATH_IMAGE048
less than threshold value
Figure 2011103318046100001DEST_PATH_IMAGE052
; Selecting its optimal mode is pattern 2, otherwise execution in step 3.3.3;
Amplitude in the step 3.3.3 compute histograms time maximum , its corresponding pattern is
Figure 2011103318046100001DEST_PATH_IMAGE056
;
Step 3.3.4 calculates poor
Figure 2011103318046100001DEST_PATH_IMAGE058
of the amplitude of two kinds of patterns; During greater than threshold value
Figure 2011103318046100001DEST_PATH_IMAGE062
, selecting the corresponding pattern
Figure 976583DEST_PATH_IMAGE050
of amplitude maximum is the optimum prediction mode of luminance block as ; Otherwise comparison pattern
Figure 371792DEST_PATH_IMAGE050
and pattern 2 corresponding matching error values, selecting the corresponding pattern of less matching error value is the optimum prediction mode of luminance block.
5. a kind of fast intra mode selection algorithm according to claim 2 is characterized in that: said threshold value
Figure 2011103318046100001DEST_PATH_IMAGE064
.
CN 201110331804 2011-10-27 2011-10-27 Fast intra-frame mode selecting algorithm Expired - Fee Related CN102387364B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 201110331804 CN102387364B (en) 2011-10-27 2011-10-27 Fast intra-frame mode selecting algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 201110331804 CN102387364B (en) 2011-10-27 2011-10-27 Fast intra-frame mode selecting algorithm

Publications (2)

Publication Number Publication Date
CN102387364A true CN102387364A (en) 2012-03-21
CN102387364B CN102387364B (en) 2013-10-23

Family

ID=45826258

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 201110331804 Expired - Fee Related CN102387364B (en) 2011-10-27 2011-10-27 Fast intra-frame mode selecting algorithm

Country Status (1)

Country Link
CN (1) CN102387364B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103686166A (en) * 2013-11-18 2014-03-26 深圳市云宙多媒体技术有限公司 Fast prediction mode selection method and system based on correlation analysis
CN106162175A (en) * 2015-03-26 2016-11-23 北京君正集成电路股份有限公司 Method for choosing frame inner forecast mode and device
CN107071405A (en) * 2016-10-27 2017-08-18 浙江大华技术股份有限公司 A kind of method for video coding and device

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2005022919A1 (en) * 2003-08-26 2005-03-10 Thomson Licensing S.A. Method and apparatus for decoding hybrid intra-inter coded blocks
CN101217663A (en) * 2008-01-09 2008-07-09 上海华平信息技术股份有限公司 A quick selecting method of the encode mode of image pixel block for the encoder
CN101309408A (en) * 2008-06-13 2008-11-19 青岛海信电子产业控股股份有限公司 Lightness block selection method of intra-frame prediction mode

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2005022919A1 (en) * 2003-08-26 2005-03-10 Thomson Licensing S.A. Method and apparatus for decoding hybrid intra-inter coded blocks
CN101217663A (en) * 2008-01-09 2008-07-09 上海华平信息技术股份有限公司 A quick selecting method of the encode mode of image pixel block for the encoder
CN101309408A (en) * 2008-06-13 2008-11-19 青岛海信电子产业控股股份有限公司 Lightness block selection method of intra-frame prediction mode

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103686166A (en) * 2013-11-18 2014-03-26 深圳市云宙多媒体技术有限公司 Fast prediction mode selection method and system based on correlation analysis
CN103686166B (en) * 2013-11-18 2017-05-17 深圳市云宙多媒体技术有限公司 Fast prediction mode selection method and system based on correlation analysis
CN106162175A (en) * 2015-03-26 2016-11-23 北京君正集成电路股份有限公司 Method for choosing frame inner forecast mode and device
CN107071405A (en) * 2016-10-27 2017-08-18 浙江大华技术股份有限公司 A kind of method for video coding and device
CN107071405B (en) * 2016-10-27 2019-09-17 浙江大华技术股份有限公司 A kind of method for video coding and device

Also Published As

Publication number Publication date
CN102387364B (en) 2013-10-23

Similar Documents

Publication Publication Date Title
CN103997646B (en) Fast intra-mode prediction mode selecting method in a kind of HD video coding
CN101964906B (en) Rapid intra-frame prediction method and device based on texture characteristics
WO2018010492A1 (en) Rapid decision making method for intra-frame prediction mode in video coding
CN103517069A (en) HEVC intra-frame prediction quick mode selection method based on texture analysis
CN105959699B (en) A kind of quick inter-frame prediction method based on estimation and time-space domain correlation
CN105208387B (en) A kind of HEVC Adaptive Mode Selection Method for Intra-Prediction
CN105141954A (en) HEVC interframe coding quick mode selection method
Shen et al. Ultra fast H. 264/AVC to HEVC transcoder
CN104052994A (en) Method for HEVC intra-frame prediction mode quick decision based on grading self-adaption
CN103327325A (en) Intra-frame prediction mode rapid self-adaptation selection method based on HEVC standard
CN103188496A (en) Fast motion estimation video encoding method based on motion vector distribution forecast
CN103596004A (en) Intra-frame prediction method and device based on mathematical statistics and classification training in HEVC
CN107343198A (en) A kind of quick decision method of AVS2 inter-frame forecast modes
CN100484249C (en) Method of forecast inside frame based on edge direction between frames
CN101304529A (en) Method and device for selecting macro block pattern
CN103384326A (en) Quick intra-frame prediction mode selection method for AVS-M video coding
CN100596202C (en) Fast mode selection method in frame
CN102364950B (en) H.264/advanced video coding (AVC)-standard-based intra-frame prediction mode rapid selection method and device
CN102075751B (en) Macro block motion state-based H264 quick mode selection method
CN114339218A (en) Image encoding method, image encoding device, electronic apparatus, and readable storage medium
CN102186081A (en) H.264 intra-frame mode selection method based on gradient vector
Huang et al. Rate-distortion-complexity optimized coding mode decision for HEVC
CN104702959B (en) A kind of intra-frame prediction method and system of Video coding
CN101867818B (en) Selection method and device of macroblock mode
CN102164283A (en) Sub-pixel motion estimation method based on AVS (audio video standard)

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20131023

Termination date: 20151027

EXPY Termination of patent right or utility model