CN107509076A - A kind of Encoding Optimization towards ultra high-definition video - Google Patents
A kind of Encoding Optimization towards ultra high-definition video Download PDFInfo
- Publication number
- CN107509076A CN107509076A CN201710740112.4A CN201710740112A CN107509076A CN 107509076 A CN107509076 A CN 107509076A CN 201710740112 A CN201710740112 A CN 201710740112A CN 107509076 A CN107509076 A CN 107509076A
- Authority
- CN
- China
- Prior art keywords
- pattern
- mode
- predicting unit
- value
- hsad
- 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
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
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/593—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving spatial prediction techniques
-
- 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
-
- 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
- H04N19/11—Selection of coding mode or of prediction mode among a plurality of spatial 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
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention discloses a kind of Encoding Optimization towards ultra high-definition video, its step includes：The size of current prediction unit is prejudged, if greater than given threshold, then utilizes step 1)~3) obtain candidate modes predicting unit is encoded；Otherwise the candidate pattern that the predicting unit is obtained using sobel operators is encoded to predicting unit；1) direction mode in predictive mode is divided into multigroup, the selection generation table schema from every group；Using non-direction mode as one group and choose generation table schema；2) thick model selection cost value computing is carried out to each pattern that represents, obtains each thick model selection cost value HSAD for representing pattern；3) some of HSAD minimums are chosen and represent pattern；To each thick model selection cost value computing of other predictive modes progress represented in the group of pattern place, obtain the HSAD of corresponding predictive mode and therefrom choose the minimum some predictive modes of HSAD as candidate modes.
Description
Technical field
It is excellent in particular to a kind of coding towards ultra high-definition video the present invention relates to computer picture coding field
Change method.
Background technology
With the development of science and technology, the arrival in large-size screen monitors epoch, the image of traditional high definition resolution ratio can not meet the people
The visual demand that crowd increasingly increases, broadcasting and TV media industry are also pursuing more perfect clearly image, are accelerating video resolution
Lifting, therefore, ultra high-definition video is increasingly becoming Hot spots for development.
Ultra high-definition video has many advantages compared with HD video, and its resolution ratio is higher, and data volume is bigger, Neng Gougeng
The good truth for going back original picture, more abundant grain details, wide visual angle and changeable color are provided to spectators, made
Spectators can obtain impression more on the spot in person.
But also make it that ultra high-definition video is faced with data transfer and data storage very big for the substantial increase of data volume
Difficulty, in order to ensure code efficiency, reduce transmission bandwidth, it is necessary to take newest coding standard to compile ultra high-definition video
Code, newest coding standard is HEVC coding standards in the world.HEVC coding standards calculate multiple compared with coding standard before
Miscellaneous degree greatly increases, and when being encoded to ultra high-definition video sequence, the required scramble time is long, and is currently based on
The optimized algorithm of HEVC coding standards is mostly towards common HD video, not well with reference to the feature of ultra high-definition video.
The content of the invention
It is a kind of it is an object of the invention to be provided based on HEVC coding standards for technical problem present in prior art
Towards the Encoding Optimization of ultra high-definition video.The size of current prediction unit is prejudged first, for larger prediction
The range of choice of intra prediction mode is reduced in the packet of unit Land use models, passes through sobel operator meters for less predicting unit
Calculate gradient information and obtain candidate pattern set, avoid the traversal of 35 kinds of predictive mode cost values, improve code efficiency, save
Scramble time.
The present invention is achieved through the following technical solutions：
If current prediction unit size is more than 8 × 8, i.e., 64 × 64 or 32 × 32 or 16 × 16, due to this several size
The pixel included in predicting unit is more, and texture is relatively scattered, and texture gradient information there may be several different directions,
So also need to carry out different predictive modes cost value calculating so as to choose less cost value, if to 35 kinds of predictive modes
Travel through the scramble time method that is more, therefore using mode packet of consuming.
The specific implementation step of mode packet is：
(A1) sequence number by this 33 kinds of direction modes of pattern 2 to pattern 34 according to pattern, one is divided per three adjacent patterns
Group, choose the representative pattern that middle model is the group, obtain 11 and represent pattern, i.e., pattern 3,6,9,12,15,18,21,24,
27、30、33；Pattern 0 and pattern 1 this 2 kinds of non-direction modes are divided into one group, because hereinafter after the pattern of representative is selected, also
Need to calculate and represent the cost values of adjacent two patterns up and down of pattern, if choosing pattern 0 to represent pattern, 0-1=-1 without
Method corresponds to corresponding mode index, so choosing pattern 1 to represent pattern.
(A2) represent pattern by 12 and carry out thick model selection (Rough Mode Decision, RMD) cost value computing：
HSAD=SATD+ λ * B
Wherein i, j represent in current prediction unit the pixel on both horizontally and vertically respectively,Represent
The size of residual values in current prediction unit between pixel original value and reconstructed value, SATD refer to predicting unit through Hadamard
Residual error size after conversion, prediction residual first carry out Hadamard conversion, and then carrying out absolute value to the residual error after conversion again asks
With.λ refers to Lagrange coefficient, and B refers to currently representing the bit number after pattern encodes predicting unit.HSAD represents meter
RMD cost values after calculation.
The formula of Hadamard conversion is as follows：
T (D)=HDH
Wherein, H is Hadamard transformation matrixs, and D is input matrix, i.e. predicting unit current pixel matrix, and its size is most
Small is 4 × 4, is up to 64 × 64.Become to bring using the Hadamard of 4 ranks and 8 ranks in HEVC and calculate SATD, the change of 4 ranks and 8 ranks
Changing matrix is respectively：
Two that minimum is selected in 12 RMD for representing pattern cost values represent Mode A, B.
(A3) other patterns represented in Mode A and B groups are carried out to the calculating of cost value, i.e. Mode A -1, A+1, B-1, B+
1, select two patterns that cost value is minimum in 6 patterns to enter RMD candidate pattern set.
If current prediction unit size is 8 × 8 or 4 × 4, due to the pixel included in the predicting unit of both sizes
Less, texture is relatively concentrated, and texture gradient information is concentrated mainly on around a certain direction, is selected by cost value computing
Candidate modes between angle degree of correlation it is very strong, candidate pattern is substantially all adjacent, the prediction residual between them
Difference is also little.So for 8 × 8,4 × 4 predicting unit, the texture gradient of sobel operator extraction images can be used to believe
Breath, so as to obtain the average gradient of current prediction unit.The grain direction angle in current prediction unit is obtained accordingly so as to obtain
Corresponding predictive mode set.
Current prediction unit gradient direction angle is obtained using sobel operators, and then obtains the candidate pattern of the predicting unit
Specific implementation step be：
(B1) it is multiplied using sobel operators with each pixel in current prediction unit, obtains the predicting unit
X, the Grad G in y directionsx(n)、Gy(n)
In formula, AnIt is for the sets of pixel values of the current prediction unit corresponding to convolution mask, n=1,2 ..., M-2, M
Current prediction unit size.Due to having neglected the pixel on edge during calculating, so the pixel actually calculated
Number is that (M-2) × (M-2) is individual.
(B2) threshold value T is set1、T2、T3、T4The gradient approximation tried to achieve in step (B1) is selected, in order to
The average gradient information of more preferable reaction current prediction unit, eliminates influence of noise.
T1=Gxmin×1.3
T2=Gxmax×0.7
T3=Gymin×1.3
T4=Gymax×0.7
G in formulaxmin、GxmaxAnd Gymin、GymaxG is represented respectivelyxAnd G (n)y(n) minimum value and maximum in.Coefficient 1.3
It is statistical value with 0.7, is finally to count to draw by many experiments.
(B3) T is chosen1<Gx(n)<T2, T3<Gy(n)<T4In the range of GxAnd G (n)y(n) value, the number of acquisition are respectively
P and q, separately constitute new gradient set G 'xAnd G ' (n)y(n)。
(B4) new gradient set G ' is tried to achievex(n) average value GxWith gradient set G 'y(n) average value Gy
Gx=(G 'x(1)+G’x(2)+...G’x(p))/p
Gy=(G 'y(1)+G’y(2)+...G’y(q))/q
(B5) the approximate grain direction angle α of current prediction unit is obtained
(B6), can be corresponding with 33 kinds of angle predictive modes according to the approximate grain direction angle α that tries to achieve, using pre- in frame
The corresponding relation of survey pattern and infra-frame prediction angle chooses the composition RMD candidate pattern set of corresponding angle predictive mode, in frame
Predictive mode and the corresponding relation of infra-frame prediction angle are as shown in table 1.
The corresponding relation of the intra prediction mode of table 1 and infra-frame prediction angle
| The scope at α angles | Candidate modes |
| [π/8,π/4] | [2]、[3]、[4]、[5] |
| [0,π/8) | [6]、[7]、[8]、[9]、[10] |
| [-π/8,0) | [10]、[11]、[12]、[13]、[14] |
| [-π/4,-π/8) | [15]、[16]、17]、[18] |
| [-3π/8,-π/4) | [18]、[19]、[20]、[21] |
| [-π/2,-3π/8) | [22]、[23]、[24]、[25]、[26] |
| [-5π/8,-π/2) | [26]、[27]、[28]、[29]、[30] |
| [-3π/4,-5π/8) | [31]、[32]、[33]、[34] |
Beneficial effects of the present invention：
The present invention applies to encode ultra high-definition video, and to ensure coding quality, the raising scramble time is target, pin
Different optimization methods is selected to different size of predicting unit, the method that mode packet is used to larger predicting unit is right
The method that less predicting unit obtains current prediction unit gradient direction using sobel operators obtains RMD candidate pattern set,
Avoid in original coding standard and 35 kinds of predictive modes are traveled through, improve code efficiency, saved the scramble time.
Specific experiment data are as shown in table 2.Scramble time in table before before representing optimizeds, after after representing optimizeds
Scramble time, BD-rate represent be the corresponding coded-bit under conditions of same PSNR compared with reference encoder efficiency
The percentage of lifting or the reduction of rate, what it was calculated is the average of the difference of two RD curves corresponding to two kinds of algorithms；BD-
What PSNR was represented is the quantity of the PSNR raisings or reduction of the reconstruction image under conditions of identical coding bit rate, and unit is
dB.The smaller explanation coding efficiencies of BD Bitrate and BD PSNR are better under normal circumstances, when BD Bitrate are less than 3, BD
When PSNR is less than 1, coding quality is can guarantee that substantially.
The experimental data of table 2
Brief description of the drawings
Fig. 1 is intra prediction mode figure in HEVC primal algorithms, wherein, pattern 2 to 34 represents 33 kinds of direction modes, uses lattice
Count to represent angle, pattern 0 and pattern 1 are non-direction mode.
Fig. 2 is to be a kind of towards ultra high-definition code optimization algorithm principle figure；
Fig. 3 is the cataloged procedure of mode packet；
(a) it is full search process, in addition to upper 33 kinds of direction modes are schemed, also two kinds of non-direction modes are respectively pattern 1
With pattern 0；
(b) it is mode packet coarse sizing, in addition to scheming upper 11 kinds of directionality and representing pattern, also a kind of non-directional represents
Pattern is pattern 1；
(c) fine screening in mode packet group.
Embodiment
For the ease of skilled artisan understands that the technology contents of the present invention, enter to present invention below in conjunction with the accompanying drawings
One step is explained.
The technical scheme is that：A kind of code optimization algorithm towards ultra high-definition video.First to ultra high-definition video
Sequence is predicted the statistics of dividing elements, and the predicting unit after HEVC primal algorithms encode to ultra high-definition sequence is drawn
Point.Predicting unit is that HEVC is carried out in frame or the elementary cell of inter prediction, its size is followed successively by 4 × 4 from small to large, 8 × 8,
16×16、32×32、64×64.The distribution situation of each predicting unit size of entire image is as shown in table 3.
Studio_dance sequences the first frame predicting unit dividing condition of table 3 counts
| Predicting unit size | 64×64 | 32×32 | 16×16 | 8×8 | 4×4 |
| Predicting unit number | 733 | 2487 | 6883 | 11072 | 4292 |
| Account for image total pixel number percentage | 36.20% | 30.70% | 21.24% | 8.54% | 0.83% |
The percentage that each predicting unit pixel count accounts for the total pixel count of entire image is calculated as follows：
Every kind of coding unit pixel count accounts for percentage=N × N × M ÷ (3840 × 2160) of image total pixel number
Wherein N represents predicting unit size, and M represents the number of predicting unit, and 3840 and 2160 be respectively ultra high-definition sequence
Pixel wide and height.
As can be seen from Table 3, when being divided with HEVC primal algorithms to 4K video sequence predicting units, bulk
Be distributed that accounting is more, about 67% pixel using 32 × 32 and the block of the above, wherein 36.20% pixel using 64 × 64 it is big
Block encodes, and less 8 × 8 and 4 × 4 block proportion only has 9.37%.Understand simultaneously, predicting unit is divided into bulk
Video image region, corresponding image texture is relatively simple, is mostly solid color regions, is divided into the video image area of fritter
Domain, corresponding image texture are comparatively complex.
Fig. 1 represents 35 kinds of intra prediction modes in HEVC coding standards, and wherein pattern 1 and pattern 0 represents two kinds of non-directions
Property predictive mode, pattern 2 to pattern 34 represent 33 kinds of directional prediction patterns.It is smooth that non-directional prediction is mainly used in texture
Region, the angle prediction of directionality is mainly used in the complicated region of texture.
Summarize and understand, for larger predicting unit, its image texture is fairly simple, and intra prediction mode is mainly non-side
The predictive mode of tropism；For less predicting unit, its image texture is complex, and intra prediction mode is mainly directionality
Predictive mode.
For the ease of understanding present context, the selection process of intra prediction direction in HEVC standard cataloged procedure is done
Go out to be expanded on further, concretely comprise the following steps：
(C1) RMD computings are carried out to all 35 predictive modes, chooses the minimum pattern of N number of RMD cost values and form one
Predictive mode set.
HSAD=SATD+ λ * B
Different predicting unit size N choosing value is also different, specific as shown in table 4：
The different predicting unit size N of table 4 choosing value
| The size of predicting unit | The N values of candidate |
| 64×64 | 3 |
| 32×32 | 3 |
| 16×16 | 3 |
| 8×8 | 8 |
| 4×4 | 8 |
(C2) the directional prediction modes information of the left side of current prediction unit and the predicting unit of top is combined, is obtained most
The set of possible predictive mode, i.e. MPM are gathered.
(C3) two set of patterns conjunction unions for obtaining step (C1) and step (C2).Calculate institute in the set after merging
There is the rate distortion costs value (Rate Distortion cost, RDcost) of predictive mode, choosing the minimum pattern of cost value is
Optimal prediction modes.
The present invention proposes a kind of code optimization algorithm towards ultra high-definition video, with reference to the predicting unit of ultra high-definition video
Relation between distribution situation and texture information, in combination with the corresponding relation between texture information and predictive mode, proposing will
The scheme that various sizes of predicting unit optimizes respectively.For larger predicting unit, due to being wrapped in this kind of predicting unit
The pixel contained is more, and texture is relatively scattered, and texture gradient information there may be several different directions, it is necessary to consider
The possibility of a variety of predictive modes, therefore using the optimization method of mode packet.For less predicting unit, because prediction is single
The pixel included in member is less, and texture is relatively concentrated, and texture gradient information is concentrated mainly on around a certain direction, therefore
The gradient information of current prediction unit is calculated using sobel operators, and then obtains gradient angle, finally gives corresponding prediction
Pattern.
Fig. 2, towards ultra high-definition code optimization algorithm principle figure, is divided into following steps to be a kind of：
(D1) judge whether current prediction unit size is more than 8 × 8, if entering in next step, walked if not jumping to
Suddenly (D4).
(D2) three points one group every to 33 kinds of direction modes, the representative pattern that middle model is the group is chosen, obtains 11
Represent pattern, i.e. pattern 3,6,9,12,15,18,21,24,27,30,33；By remaining 2 patterns, i.e. { 0,1 } after packet
It is considered as one group, selection represents pattern as 1；Pattern { 1,3,6,9,12,15,18,21,24,27,30,33 } is represented by 12 to carry out
RMD computings, choose RMD cost values minimum two represent Mode A, B.
(D3) RMD computings, i.e. A-1, A+1, B-1, B+1 are carried out to A, the two predictive modes represented in modal sets of B.Choosing
Take two that cost value is minimum in 6 patterns and enter RMD candidate pattern set.Skip to step (D5).
(D4) Grad is obtained with the image texture information of sobel operator extraction current prediction units, according to threshold value to institute
Obtained gradient approximation is selected, then the Grad in threshold value is each averaging, and obtains corresponding tangent angle, the angle
The as grain direction of current prediction unit, corresponding candidate modes are obtained further according to the corresponding relation in table 1.
(D5) most probable predictive mode set MPM computings are carried out.
(D6) MPM patterns and RMD candidate patterns are subjected to RDcost computings.
(D7) it is optimal direction predictive mode to take the minimum pattern of RDcost cost values.
It is more than 8 × 8 predicting unit for size, using the optimization process of mode packet as shown in figure 3, specific steps
For：
(E1) by every three points one group of 33 kinds of direction modes, the representative pattern that middle model is the group is chosen, obtains 11
Represent pattern, i.e. pattern 3,6,9,12,15,18,21,24,27,30,33.By 2 kinds of non-direction modes, pattern 1 and pattern 0, which are divided, is
One group, selection pattern 1 is to represent pattern.
(E2) represent pattern by 12 and carry out RMD cost value computings：
HSAD=SATD+ λ * B
Wherein SATD refers to the residual error size after conversion, and prediction residual first carries out Hadamard conversion, then again to conversion
Residual error afterwards carries out absolute value summation.λ refers to Lagrange coefficient, and B refers to the bit number after present mode coding.HSAD
Represent the RMD cost values after calculating.
The formula of Hadamard conversion is as follows：
T (D)=HDH
Wherein, H is Hadamard transformation matrixs, and D is input matrix.The Hadamard of 4 ranks and 8 ranks conversion is used in HEVC
To calculate SATD, the transformation matrix of 4 ranks and 8 ranks is respectively：
Two that minimum is selected in 12 RMD for representing pattern cost values represent Mode A, B
(E3) other patterns represented in Mode A and B groups are carried out to the calculating of cost value, i.e. Mode A -1, A+1, B-1, B+
1, select two patterns that cost value is minimum in 6 patterns to enter RMD candidate pattern set.
For the predicting unit that size is 4 × 4,8 × 8, using the optimized algorithm of sobel operator extraction gradient informations, tool
Body step is：
(F1) it is multiplied using sobel operators with each pixel in current prediction unit, obtains Grad Gx
(n)、Gy(n)
In formula, AnIt is for the sets of pixel values of the current prediction unit corresponding to convolution mask, n=1,2 ..., M-2, M
Current prediction unit size.Due to having neglected the pixel on edge during calculating, so the pixel actually calculated
Number is that (M-2) × (M-2) is individual.
(F2) threshold value T is set1、T2、T 3、T4, the gradient approximation tried to achieve in step (F1) is selected, it is therefore an objective to be
The average gradient information of more preferable reaction current prediction unit, eliminates influence of noise.
T1=Gxmin×1.3
T2=Gxmax×0.7
T3=Gymin×1.3
T4=Gymax×0.7
G in formulaxmin、GxmaxAnd Gymin、GymaxG is represented respectivelyxAnd G (n)y(n) minimum value and maximum in.Coefficient 1.3
It is statistical value with 0.7, is finally to count to draw by many experiments.
(F3) T is chosen1<Gx(n)<T2, T3<Gy(n)<T4In the range of GxAnd G (n)y(n) value, the number of acquisition are respectively
P and q, separately constitute new gradient set G 'xAnd G ' (n)y(n)。
(F4) new gradient set G ' is tried to achievex(n) average value GxWith gradient set G 'y(n) average value Gy：
Gx=(G 'x(1)+G’x(2)+...G’x(p))/p；
Gy=(G 'y(1)+G’y(2)+...G’y(q))/q。
(F5) the approximate grain direction angle of current prediction unit is obtained
(F6), can be corresponding with 33 kinds of angle predictive modes according to the approximate grain direction angle α that tries to achieve, using pre- in frame
The corresponding relation of survey pattern and infra-frame prediction angle chooses the composition RMD candidate pattern set of corresponding angle predictive mode, in frame
Predictive mode and the corresponding relation of infra-frame prediction angle are as shown in table 5.
Table 5 is intra prediction mode and the mapping table of infra-frame prediction angle
| The scope at α angles | Candidate modes |
| [π/8,π/4] | [2]、[3]、[4]、[5] |
| [0,π/8) | [6]、[7]、[8]、[9]、[10] |
| [-π/8,0) | [10]、[11]、[12]、[13]、[14] |
| [-π/4,-π/8) | [15]、[16]、17]、[18] |
| [-3π/8,-π/4) | [18]、[19]、[20]、[21] |
| [-π/2,-3π/8) | [22]、[23]、[24]、[25]、[26] |
| [-5π/8,-π/2) | [26]、[27]、[28]、[29]、[30] |
| [-3π/4,-5π/8) | [31]、[32]、[33]、[34] |
Obtain through the above way the set of RMD predictive modes and then with MPM predictive mode collection conjunction unions, and to collection
All predictive modes carry out RDcost calculating in conjunction, and it is optimal to choose the minimum predictive mode of cost value.Inventive algorithm subtracts
The number of computations of RMD cost values is lacked, binding pattern packet avoids time of 35 kinds of model prediction patterns with texture gradient information
Go through, while gather the number of inner estimation mode, RDcost number of computations after again reducing, so as to save by reducing RMD
The scramble time is saved.Through experimental test, this algorithm is for 4K video sequences on the premise of coding quality is ensured, average reduction is about
30% scramble time.
The explanation of the preferred embodiment of the present invention contained above, this be in order to describe the technical characteristic of the present invention in detail, and
It is not intended to the content of the invention being limited in the concrete form described by embodiment, according to other of present invention purport progress
Modifications and variations are also protected by this patent.The purport of present invention is to be defined by the claims, rather than has embodiment
Specific descriptions are defined.
Claims (7)
1. a kind of Encoding Optimization towards ultra high-definition video, its step includes：
The size of current prediction unit is prejudged, if greater than given threshold, then utilizes step 1)~3) to obtain candidate pre-
Survey pattern encodes to the predicting unit；Otherwise the candidate pattern of the predicting unit is obtained to the prediction using sobel operators
Unit is encoded；
1) direction mode in predictive mode is divided into it is multigroup, chosen from every group a direction mode be the group representative pattern；
Using non-direction mode as one group and choose a wherein pattern as non-direction mode group representative pattern；
2) thick model selection cost value computing is carried out to each pattern that represents, obtains each roughcast formula choosing for representing pattern
Select cost value HSAD；
3) thick some of model selection cost value HSAD minimums are chosen and represent pattern；Other in group where pattern are represented to each
Predictive mode carries out thick model selection cost value computing, obtains the thick model selection cost value HSAD and therefrom of corresponding predictive mode
The minimum some predictive modes of thick model selection cost value HSAD are chosen as candidate modes.
2. Encoding Optimization as claimed in claim 1, it is characterised in that be divided into the direction mode in predictive mode multigroup
Method be：One group will be divided into per three adjacent direction modes.
3. Encoding Optimization as claimed in claim 1 or 2, it is characterised in that choose the direction mode of the centre in every group
Representative pattern as the group.
4. Encoding Optimization as claimed in claim 1 or 2, in the step 3), thick model selection cost value HSAD is chosen
Two minimum predictive modes are used as and represent Mode A, B；Choose two minimum predictive mode conducts of thick model selection cost value HSAD
Candidate modes.
5. Encoding Optimization as claimed in claim 1, it is characterised in that described that the prediction list is obtained using sobel operators
The method of candidate pattern of member is：
1) it is multiplied using sobel operators with each pixel in the predicting unit, obtains the ladder in predicting unit x directions
Angle value Gx(n), the Grad G in predicting unit y directionsy(n)；
2) threshold value T is set1、T2、T3、T4；Choose T1<Gx(n)<T2In the range of Gx(n) value, new gradient set G ' is formedx
(n)；Choose T3<Gy(n)<T4In the range of Gy(n) value, new gradient set G ' is formedxAnd G ' (n)y(n)；Wherein, gradient collection
Close G 'x(n) element number in is p, gradient set G 'y(n) element number in is q；
3) gradient set G ' is calculatedx(n) average value GxWith gradient set G 'y(n) average value Gy；
4) the approximate grain direction angle of the predicting unit is calculated
5) according to the approximate grain direction angle α, and intra prediction mode and the corresponding relation of infra-frame prediction angle, it is pre- to obtain this
Survey the candidate pattern of unit.
6. Encoding Optimization as claimed in claim 5, it is characterised in that T1=Gxmin×1.3、T2=Gxmax×0.7、T3=
Gymin×1.3、T4=Gymax×0.7；Gxmin、GxmaxG is represented respectivelyx(n) minimum value and maximum in, Gymin、GymaxTable respectively
Show Gy(n) minimum value and maximum in.
7. Encoding Optimization as claimed in claim 1, it is characterised in that the given threshold is 8 × 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710740112.4A CN107509076B (en) | 2017-08-25 | 2017-08-25 | A kind of Encoding Optimization towards ultra high-definition video |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710740112.4A CN107509076B (en) | 2017-08-25 | 2017-08-25 | A kind of Encoding Optimization towards ultra high-definition video |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107509076A true CN107509076A (en) | 2017-12-22 |
| CN107509076B CN107509076B (en) | 2019-10-18 |
Family
ID=60693500
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710740112.4A Active CN107509076B (en) | 2017-08-25 | 2017-08-25 | A kind of Encoding Optimization towards ultra high-definition video |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107509076B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109302605A (en) * | 2018-10-26 | 2019-02-01 | 西安科锐盛创新科技有限公司 | Image encoding method and equipment based on multi-core processor |
| CN109413421A (en) * | 2018-10-26 | 2019-03-01 | 西安科锐盛创新科技有限公司 | Method for video coding, equipment and video encoding/decoding method, equipment |
| CN109462758A (en) * | 2018-10-09 | 2019-03-12 | 深圳市网心科技有限公司 | Intra prediction mode determines method, electronic equipment, system and storage medium |
| CN109561303A (en) * | 2018-10-26 | 2019-04-02 | 西安科锐盛创新科技有限公司 | A kind of prediction technique based on video compress |
| CN109587494A (en) * | 2018-10-26 | 2019-04-05 | 西安科锐盛创新科技有限公司 | The prediction technique of video compress |
| CN109587481A (en) * | 2018-10-26 | 2019-04-05 | 西安科锐盛创新科技有限公司 | Method for video coding and equipment |
| CN109660793A (en) * | 2018-10-26 | 2019-04-19 | 西安科锐盛创新科技有限公司 | Prediction technique for bandwidth reduction |
| CN110213576A (en) * | 2018-05-03 | 2019-09-06 | 腾讯科技（深圳）有限公司 | Method for video coding, video coding apparatus, electronic equipment and storage medium |
| CN110446048A (en) * | 2018-05-04 | 2019-11-12 | 阿里巴巴集团控股有限公司 | The choosing method and device of prediction mode in intra prediction scene |
| CN112752105A (en) * | 2021-01-04 | 2021-05-04 | 暨南大学 | Intra-frame prediction coding complexity reduction method suitable for HEVC standard |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130266069A1 (en) * | 2010-12-17 | 2013-10-10 | Toshiyasu Sugio | Image coding method and image decoding method |
| JP2014225795A (en) * | 2013-05-16 | 2014-12-04 | 日本電信電話株式会社 | Intra-prediction direction refining method, intra-prediction direction refining device and intra-prediction direction refining program |
| CN104284186A (en) * | 2014-09-24 | 2015-01-14 | 复旦大学 | Fast algorithm suitable for HEVC standard intra-frame prediction mode judgment process |
| CN105187826A (en) * | 2015-07-31 | 2015-12-23 | 郑州轻工业学院 | Rapid intra-frame mode decision method specific to high efficiency video coding standard |
-
2017
- 2017-08-25 CN CN201710740112.4A patent/CN107509076B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130266069A1 (en) * | 2010-12-17 | 2013-10-10 | Toshiyasu Sugio | Image coding method and image decoding method |
| JP2014225795A (en) * | 2013-05-16 | 2014-12-04 | 日本電信電話株式会社 | Intra-prediction direction refining method, intra-prediction direction refining device and intra-prediction direction refining program |
| CN104284186A (en) * | 2014-09-24 | 2015-01-14 | 复旦大学 | Fast algorithm suitable for HEVC standard intra-frame prediction mode judgment process |
| CN105187826A (en) * | 2015-07-31 | 2015-12-23 | 郑州轻工业学院 | Rapid intra-frame mode decision method specific to high efficiency video coding standard |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110213576A (en) * | 2018-05-03 | 2019-09-06 | 腾讯科技（深圳）有限公司 | Method for video coding, video coding apparatus, electronic equipment and storage medium |
| CN110213576B (en) * | 2018-05-03 | 2023-02-28 | 腾讯科技（深圳）有限公司 | Video encoding method, video encoding device, electronic device, and storage medium |
| CN110446048A (en) * | 2018-05-04 | 2019-11-12 | 阿里巴巴集团控股有限公司 | The choosing method and device of prediction mode in intra prediction scene |
| CN109462758A (en) * | 2018-10-09 | 2019-03-12 | 深圳市网心科技有限公司 | Intra prediction mode determines method, electronic equipment, system and storage medium |
| CN109462758B (en) * | 2018-10-09 | 2021-04-06 | 深圳市网心科技有限公司 | Intra-frame prediction mode determination method, electronic device, system and storage medium |
| CN109587494A (en) * | 2018-10-26 | 2019-04-05 | 西安科锐盛创新科技有限公司 | The prediction technique of video compress |
| CN109660793A (en) * | 2018-10-26 | 2019-04-19 | 西安科锐盛创新科技有限公司 | Prediction technique for bandwidth reduction |
| CN109587481A (en) * | 2018-10-26 | 2019-04-05 | 西安科锐盛创新科技有限公司 | Method for video coding and equipment |
| CN109302605A (en) * | 2018-10-26 | 2019-02-01 | 西安科锐盛创新科技有限公司 | Image encoding method and equipment based on multi-core processor |
| CN109587494B (en) * | 2018-10-26 | 2020-07-10 | 江苏怀业信息技术股份有限公司 | Prediction method for video compression |
| CN109587481B (en) * | 2018-10-26 | 2020-07-10 | 江苏怀业信息技术股份有限公司 | Video encoding method and apparatus |
| CN109561303B (en) * | 2018-10-26 | 2020-12-15 | 广东弘视数字传媒有限公司 | Prediction method based on video compression |
| CN109413421B (en) * | 2018-10-26 | 2021-01-19 | 张豪 | Video encoding method, video encoding apparatus, video decoding method, and video decoding apparatus |
| CN109561303A (en) * | 2018-10-26 | 2019-04-02 | 西安科锐盛创新科技有限公司 | A kind of prediction technique based on video compress |
| CN109413421A (en) * | 2018-10-26 | 2019-03-01 | 西安科锐盛创新科技有限公司 | Method for video coding, equipment and video encoding/decoding method, equipment |
| CN112752105A (en) * | 2021-01-04 | 2021-05-04 | 暨南大学 | Intra-frame prediction coding complexity reduction method suitable for HEVC standard |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107509076B (en) | 2019-10-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107509076B (en) | A kind of Encoding Optimization towards ultra high-definition video | |
| CN103220527B (en) | Method for encoding images and device and its coding/decoding method and device | |
| CN105208387B (en) | A kind of HEVC Adaptive Mode Selection Method for Intra-Prediction | |
| CN108322747A (en) | A kind of coding unit Partitioning optimization method towards ultra high-definition video | |
| CN104918056B (en) | The method decoded to intra prediction mode | |
| CN101964906B (en) | Rapid intra-frame prediction method and device based on texture characteristics | |
| CN104539962A (en) | Layered video coding method fused with visual perception features | |
| CN107071416A (en) | A kind of HEVC Adaptive Mode Selection Method for Intra-Prediction | |
| CN106534846A (en) | Method for dividing and quickly encoding screen contents and natural contents | |
| US20120039389A1 (en) | Distortion weighing | |
| CN109068142B (en) | 360-degree video intra-frame prediction rapid decision-making method, device, coder-decoder and medium | |
| CN107864380B (en) | 3D-HEVC fast intra-frame prediction decision method based on DCT | |
| CN105791862B (en) | 3 d video encoding depth map internal schema selection method based on fringe complexity | |
| CN107197267B (en) | The distribution method of efficient video coding criterion and quantity parameter | |
| CN106534862A (en) | Video coding method | |
| US20030123550A1 (en) | Fast motion estimation using N-queen pixel decimation | |
| CN106688238A (en) | Improved reference pixel selection and filtering for intra coding of depth map | |
| CN106303530A (en) | A kind of bit rate control method merging vision perception characteristic | |
| CN112188196A (en) | Method for rapid intra-frame prediction of general video coding based on texture | |
| CN102917226B (en) | Intra-frame video coding method based on self-adaption downsampling and interpolation | |
| CN103313055B (en) | A kind of chroma intra prediction method based on segmentation and video code and decode method | |
| CN114885172A (en) | Fast coding unit dividing method for H.266/VVC intra-frame prediction coding | |
| CN114040211A (en) | AVS 3-based intra-frame prediction rapid decision-making method | |
| CN106878754A (en) | A kind of 3D video depths image method for choosing frame inner forecast mode | |
| CN1130920C (en) | Mode coding method in binary shape encoding |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |