CN103561265A - Method for integer DST conversion base selection for video coding - Google Patents

Method for integer DST conversion base selection for video coding Download PDF

Info

Publication number
CN103561265A
CN103561265A CN201310573819.2A CN201310573819A CN103561265A CN 103561265 A CN103561265 A CN 103561265A CN 201310573819 A CN201310573819 A CN 201310573819A CN 103561265 A CN103561265 A CN 103561265A
Authority
CN
China
Prior art keywords
integer
coefficient
matrix
transform
search
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
Application number
CN201310573819.2A
Other languages
Chinese (zh)
Inventor
王中元
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.)
Wuhan University WHU
Original Assignee
Wuhan University WHU
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 Wuhan University WHU filed Critical Wuhan University WHU
Priority to CN201310573819.2A priority Critical patent/CN103561265A/en
Publication of CN103561265A publication Critical patent/CN103561265A/en
Pending legal-status Critical Current

Links

Abstract

The invention discloses a method for integer DST conversion base selection for video coding. The method comprises the following steps that step 1, according to a DST conversion coefficient matrix C, an expression of an integer conversion matrix I corresponding to the DST conversion coefficient matrix C is obtained through calculation, wherein the coefficient of the C is marked to be ck, the coefficient of the I is marked to be ik, and k=0-N; step 2, ik is normalized so that a calibration coefficient jk can be obtained, wherein k=0-N; step 3, ik is searched for and calculated to form a conversion base through a searching method and the process of searching meets the following constraint conditions that elements except for diagonals in an IIT matrix all meet the condition that the values are zero;, the step size in search is limited to be 1/2 when ik is searched for and obtained through calculation, the calibration coefficient jk needs to satisfy the requirement of a threshold value which is decided by the number of conversion bases expected to be obtained through the search, and the amplitude peak of the coefficient ik in the obtained integer conversion matrix I is 15; step 4, the conversion bases obtained through the search are standardized to integers and the integer conversion base corresponding to the matrix C is obtained. According to the method, the integer DST conversion base with the high quality can be obtained completely and rapidly.

Description

A kind of integer DST transform-based system of selection of Video coding
Technical field
The invention belongs to field of video encoding, particularly the method for optimizing of the integer DST transform-based in a kind of video encoding standard and the transform-based preferably obtaining.
Background technology
In April, 2013, International Telecommunication Association (ITU) has formally issued HEVC/H.265 standard, and standard full name is efficient video coding (High Efficiency Video Coding).H.265, H.264 standard compared to before, has had sizable improvement, only needs original half code check can play the video of equal in quality.H.265 standard has adopted the multinomial new technology relation between balanced code flow structure, coding quality, time delay and algorithm complex in addition.Wherein, aspect conversion, H.265 first discrete sine transform (DST) is incorporated in standard.Before this, the transform method dominating in video encoding standard technology is discrete cosine transform (DCT) always.
DST conversion has effectively improved the transition coding efficiency of intra-frame prediction block.Infra-frame prediction often relatively the neighbor pixel of having encoded of the image block left side or top launch, therefore, the pixel from neighbours close to more obtains higher precision of prediction less prediction residual in other words than the pixel energy away from neighbours.Quantity research shows greatly, and with respect to dct transform, DST conversion is more suitable for describing this statistical property of intra-frame prediction block.Therefore, H.265, when retaining existing dct transform, formally DST conversion is brought in standard, be specifically designed to the 4x4 transition coding of infra-frame prediction luminance block.And, in order to reduce mapping fault, improve arithmetic speed, absorbed the H.264 thought of middle integer transform, adopted the implementation without multiplication integer transform of approximate DST.
Be different from common DST conversion, integer DST transform-based is not unique.Searching GL is strong, the simple integer translation base of computing is an important job in video encoding standard process.H.265 the integer translation base in has obtained patent protection, and China is stepping up the native country standard A VS2 formulating with H.265 performance is suitable, therefore, must propose the new integer translation base having adopted in being different from H.265.
Summary of the invention
In order fundamentally to have solved pinpointing the problems of transform-based, the present invention is based on the principle of integer DST conversion, a kind of general generating algorithm of transform-based has been proposed, this invention has positive role for the video compression standard of formulating China's independent intellectual property right.
The technical solution adopted in the present invention is: a kind of integer DST transform-based system of selection of Video coding, it is characterized in that, and comprise the following steps:
Step 1: according to DST transform coefficient matrix C, solve the expression formula of its corresponding integer transform matrix I, wherein the coefficient of C is labeled as c k, the coefficient of I is labeled as i k, k=0,1 ..., N; Wherein, N represents the number of the coefficient that numerical value in DST transform coefficient matrix is different;
Step 2: by i knormalization, tries to achieve calibration coefficient j k, k=0,1...,N;
Step 3: adopt searching method search finding i kto form transform-based, wherein, according to following 4 constraintss, carry out search finding i k,
Constraints one: II tit is 0 that element in matrix beyond diagonal all meets value, described II tmatrix is multiplied by its transposed matrix by integer transform matrix I and obtains;
Constraints two: restriction search finding i ktime step-size in search be 1/2;
Constraints three: calibration coefficient j kneed to meet &Sigma; k = 0 N | j k - c k | < T f , k = 0,1 , . . . , N , T frepresent thresholding, threshold value depends on the transform-based number that expectation search obtains;
Constraints four: coefficient i in the round numbers transformation matrix I of institute kamplitude peak be 15;
Step 4: the transform-based yardstickization that search is obtained, to integer, obtains integer translation base corresponding to DST transform coefficient matrix C.
As preferably, described calibration coefficient j kneed to meet &Sigma; k = 0 N | j k - c k | * f k < T f , k = 0,1 , . . . , N , Weight coefficient wherein f k = c k &Sum; j = 0 j = N c j , k = 0,1 . . . N .
As preferably, described T fvalue be 0.03.
As preferably, utilize the integer DST transform-based system of selection of described Video coding, 2 coefficients, 0.371748,0.601501 search in 4x4DST transformation matrix is obtained to 10 groups of corresponding integer translation bases, and 2 coefficients in its 10 groups of integer translation bases are respectively:
(1)4、6.5;
(2)2.5、4;
(3)1.5、2.5;
(4)3.5、5.5;
(5)3.5、6;
(6)2、3.5;
(7)1、1.5;
(8)2.5、4.5;
(9)4.5、6.5;
(10)3、5.5。
As preferably, 10 groups of described integer translation bases are with the deviation of 2 coefficients in original 4x4DST transformation matrix according to ascending arrangement, and the base that deviation is less is better to the reservation of original 4x4DST character, preferentially selects in use.
The present invention is by analyzing the principle of integer DST conversion, four constraintss that integer transform matrix should meet have been pointed out, then to meet orthogonality, be constrained to starting point, derived the quantitative relation between integer transform matrix element, and be aided with other three constraintss, adopt searching method to find integer translation base.Adopt technical scheme method of the present invention, only need after the search of tens steps, just can find out all available integer translation bases, efficiency is high and comprehensive.The present invention is not only conducive to improve the efficiency of DST transition coding, for China, formulates and has the video compression standard of independent intellectual property right significant.
Embodiment
Below with reference to specific embodiment, the present invention is further elaborated.
The technical solution adopted in the present invention is: a kind of integer DST transform-based system of selection of Video coding, comprises the following steps:
Step 1: according to DST transform coefficient matrix C, solve the expression formula of its corresponding integer transform matrix I, wherein the coefficient of C is labeled as c k, the coefficient of I is labeled as i k, k=0,1 ..., N; DST transform coefficient matrix C generally comprises M * M element, M=4,8,16 N represents the number of the coefficient that numerical value in DST transform coefficient matrix is different, and numerical value is determined as the case may be;
Step 2: by i knormalization, tries to achieve calibration coefficient j k, k=0,1 ..., N;
Step 3: adopt searching method search finding i kto form transform-based, wherein, according to following 4 constraintss, carry out search finding i k,
Constraints one: II tit is 0 that element in matrix beyond diagonal all meets value, described II tmatrix is multiplied by its transposed matrix by integer transform matrix I and obtains;
Constraints two: restriction search finding i ktime step-size in search be 1/2;
Constraints three: calibration coefficient j kneed to meet &Sigma; k = 0 N | j k - c k | < T f , k = 0,1 , . . . , N , T frepresent thresholding, threshold value depends on the transform-based number that expectation search obtains; For the constraints of further strictly searching for, calibration coefficient j kneed to meet &Sigma; k = 0 N | j k - c k | * f k < T f , k = 0,1 , . . . , N , Weight coefficient wherein f k = c k &Sum; j = 0 j = N c j , k = 0,1 . . . N ; T fvalue be 0.03;
Constraints four: coefficient i in the round numbers transformation matrix I of institute kamplitude peak be 15;
Step 4: the transform-based yardstickization that search is obtained, to integer, obtains integer translation base corresponding to DST transform coefficient matrix C.
Utilize the integer DST transform-based system of selection of Video coding, 2 coefficients, 0.371748,0.601501 search in 4x4DST transformation matrix is obtained to 10 groups of corresponding integer translation bases, 2 coefficients in its 10 groups of integer translation bases are respectively:
(1)4、6.5;
(2)2.5、4;
(3)1.5、2.5;
(4)3.5、5.5;
(5)3.5、6;
(6)2、3.5;
(7)1、1.5;
(8)2.5、4.5;
(9)4.5、6.5;
(10)3、5.5。
10 groups of above-mentioned integer translation bases are with the deviation of 2 coefficients in original 4x4DST transformation matrix according to ascending arrangement, and the base that deviation is less is better to the reservation of original 4x4DST character, preferentially select in use.
Due to integer transform computing, only allow to occur the arithmetical operations such as adding, subtract, taking advantage of of integer type and displacement, so the coefficient i that search obtains kexcept integer, can also there is 0.5 such value (0.5 means the computing of 1 of moving to right).Can be by the calibration coefficient j obtaining in search procedure during concrete enforcement kform normalization matrix, by revising normalization matrix, transform-based can yardstickization arrive integer.
4 fundamental characteristics of integer transform matrix are: 1. can be normalized into orthogonal matrix; 2. transform operation can all be realized with shift operation; 3. the orthogonal matrix after normalization is close with original DST transformation matrix coefficient; 4. integer transform matrix coefficient is not very large, and transform operation can be with 16 accuracy representings.The present invention is according to above-mentioned 4 principles, and the method for optimizing of design integer transform matrix, is converted to 4 fundamental characteristics the constraints of searching for transform-based, obtains suitable integer translation base under effective search comprehensively.Wherein the corresponding characteristic of constraints one 1.; The corresponding characteristic of constraints two 2.; The corresponding characteristic of constraints three 3.; The corresponding characteristic of constraints four 4..Because 1. characteristic is basis, during search, should be constrained to starting point to meet orthogonality, derived the quantitative relation between integer transform matrix I interior element, be then aided with other three constraintss and adopt searching method to find transform-based.Owing to having provided fully preferably constraints, in tens step search, just can find certain coefficient c kcorresponding all preferred integer translation base.
Thresholding T during concrete enforcement fnumerical value depends on the transform-based number that expectation search obtains, and can increase its numerical value, take to reduce transform-based quality and obtain more candidates' transform-based as cost, and the preferred embodiment of the present invention adopts 0.03.Consideration characteristics is 3. time, and the contribute energy of the larger coefficient of the amplitude that it is also conceivable that after to conversion is larger, further strict search constraints condition three calibration coefficient j kneed meet and be &Sigma; k = 0 N | j k - c k | * f k < T f , k = 0,1 , . . . , N , Weight coefficient wherein after weight coefficient correction, ask for the integer translation base obtaining more accurate.
Next with the embodiment that is solved to of 4x4 integer DST transformation matrix, its generative process is as follows:
According to transformation kernel matrix T i , k = 2 N + 1 sin [ &pi; ( i + 1 ) ( k + 1 ) N + 1 ] Obtaining 4x4DST transform coefficient matrix is
0.371748 0.601501 0.601501 0.371748 0.601501 0.371748 - 0.371748 - 0.601501 0.601501 - 0.371748 - 0.371748 0.601501 0.371748 - 0.601501 0.601501 - 0.371748
Do not consider symbol, coefficient has 2 kinds of values: 0.371748,0.601501, be designated as successively: c 0, c 1, make the integer DST transformation matrix coefficient of their correspondences be respectively: i 0, i 1, INTEGER MATRICES I can be written as
i 0 i 1 i 1 i 0 i 1 i 0 - i 0 - i 1 i 1 - i 0 - i 0 i 0 i 0 - i 1 i 1 - i 0
In order to meet orthogonality, consider
II T = a 0 0 0 0 a 0 0 0 0 a 0 0 0 0 a
Wherein,
a=2(i 0 2+i 1 2)
By II texpression-form known, 1. character be met.In order to meet character 2., it is 1/2 that the present invention gets step-size in search, and for meeting 4., the amplitude peak of getting coefficient is 15; For meeting character 3., by coefficient i 0, i 1normalization obtains calibration coefficient j 0, j 1,
j k = i k a k = 0,1
Therefore, 3. character can be expressed as
&Sigma; k = 0 1 | j k - c k | < T , k = 0,1
Here T represents precision control thresholding.Consider that the contribute energy after large coefficient is to conversion is larger, therefore weighting, weighted factor is defined as
f k = c k &Sum; j = 0 j = 1 c j , k = 0,1
So above formula is further write as
&Sigma; k = 0 1 | j k - c k | * f k < T f , k = 0,1
Thresholding T fthe candidate transformation radix order that depends on expectation, the present embodiment gets 0.03.
According to coefficient after demarcating, with the deviation of original DST coefficient (0.371748,0.601501) order from small to large, table 1 has been listed front 10 groups of transform-based.The base that deviation is less is better to the reservation of original DST character, should preferentially select in use.It is pointed out that, by revising normalization matrix G, transform-based can yardstickization arrive integer.
Table 14x4 integer DST transform-based
Figure BDA0000414949140000071
The candidate transformation basic group that provides in table 1 belongs to and not yet obtains in prior art; when designing new video compression standard, can select newfound transform-based; thereby the patent obstacle of evading offshore company; for China to occupy superiority aspect international video encoding standard significant, the present invention's protection that claims especially.
These are only preferred embodiment of the present invention, be not intended to limit protection scope of the present invention, therefore, all any modifications of doing within the spirit and principles in the present invention, be equal to replacement, improvement etc., within all should being included in protection scope of the present invention.

Claims (5)

1. an integer DST transform-based system of selection for Video coding, is characterized in that, comprises the following steps:
Step 1: according to DST transform coefficient matrix C, solve the expression formula of its corresponding integer transform matrix I, wherein the coefficient of C is labeled as c k, the coefficient of I is labeled as i k, k=0,1 ..., N; Wherein, N represents the number of the coefficient that numerical value in DST transform coefficient matrix is different;
Step 2: by i knormalization, tries to achieve calibration coefficient j k, k=0,1 ..., N;
Step 3: adopt searching method search finding i kto form transform-based, wherein, according to following 4 constraintss, carry out search finding i k,
Constraints one: II tit is 0 that element in matrix beyond diagonal all meets value, described II tmatrix is multiplied by its transposed matrix by integer transform matrix I and obtains;
Constraints two: restriction search finding i ktime step-size in search be 1/2;
Constraints three: calibration coefficient j kneed to meet &Sigma; k = 0 N | j k - c k | < T f , k = 0,1 , . . . , N , T frepresent thresholding, threshold value depends on the transform-based number that expectation search obtains;
Constraints four: coefficient i in the round numbers transformation matrix I of institute kamplitude peak be 15;
Step 4: the transform-based yardstickization that search is obtained, to integer, obtains integer translation base corresponding to DST transform coefficient matrix C.
2. the integer DST transform-based system of selection of Video coding according to claim 1, is characterized in that: described calibration coefficient j kneed to meet &Sigma; k = 0 N | j k - c k | * f k < T f , k = 0,1 , . . . , N , Weight coefficient wherein
f k = c k &Sum; j = 0 j = N c j , k = 0,1 . . . N .
3. the integer DST transform-based system of selection of Video coding according to claim 1 and 2, is characterized in that: described T fvalue be 0.03.
4. the integer DST transform-based system of selection of Video coding according to claim 1, it is characterized in that: utilize the integer DST transform-based system of selection of described Video coding, 2 coefficients, 0.371748,0.601501 search in 4x4DST transformation matrix is obtained to 10 groups of corresponding integer translation bases, and 2 coefficients in its 10 groups of integer translation bases are respectively:
(1)4、6.5;
(2)2.5、4;
(3)1.5、2.5;
(4)3.5、5.5;
(5)3.5、6;
(6)2、3.5;
(7)1、1.5;
(8)2.5、4.5;
(9)4.5、6.5;
(10)3、5.5。
5. the integer DST transform-based system of selection of Video coding according to claim 4, it is characterized in that: 10 groups of described integer translation bases with the deviation of 2 coefficients in original 4x4DST transformation matrix according to ascending arrangement, the base that deviation is less is better to the reservation of original 4x4DST character, preferentially selects in use.
CN201310573819.2A 2013-11-15 2013-11-15 Method for integer DST conversion base selection for video coding Pending CN103561265A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201310573819.2A CN103561265A (en) 2013-11-15 2013-11-15 Method for integer DST conversion base selection for video coding

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201310573819.2A CN103561265A (en) 2013-11-15 2013-11-15 Method for integer DST conversion base selection for video coding

Publications (1)

Publication Number Publication Date
CN103561265A true CN103561265A (en) 2014-02-05

Family

ID=50015402

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201310573819.2A Pending CN103561265A (en) 2013-11-15 2013-11-15 Method for integer DST conversion base selection for video coding

Country Status (1)

Country Link
CN (1) CN103561265A (en)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1564602A (en) * 2004-03-18 2005-01-12 华中科技大学 Integral conversing matrix selection method of video coding and related integral conversion method
CN101198059A (en) * 2007-12-27 2008-06-11 武汉大学 Integer translation base optimization method in video coding standard

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1564602A (en) * 2004-03-18 2005-01-12 华中科技大学 Integral conversing matrix selection method of video coding and related integral conversion method
CN101198059A (en) * 2007-12-27 2008-06-11 武汉大学 Integer translation base optimization method in video coding standard

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
王中元等: "整数DCT变换中变换基的通用生成算法", 《中国图像图形学报》 *
王中元等: "视频编码器中DCT域系数的变换编码研究", 《通信学报》 *

Similar Documents

Publication Publication Date Title
Huang et al. Shuffle transformer: Rethinking spatial shuffle for vision transformer
CN105898330B (en) The method and apparatus that Video coding is carried out using limited frame intra block copy mode
CN103473307B (en) Across media sparse hash indexing means
CN105338351B (en) Intra prediction coding and decoding, array scanning method and device based on template matching
CN107027025B (en) A kind of light field image compression method based on macro block of pixels adaptive prediction
Du et al. Adaptive sparse convolutional networks with global context enhancement for faster object detection on drone images
CN101917622B (en) 14-bit width image compression hardware coder
CN105653640A (en) Collaborative filtering recommendation method based on trust mechanism
CN100592795C (en) Integer translation base optimization method in video coding standard
CN112487020B (en) Method and system for converting graph of SQL to text into natural language statement
CN103873864A (en) Object flag bit efficient encoding method applied to video object retrieval
CN109934272A (en) A kind of image matching method based on full convolutional network
CN104202605B (en) A kind of method and its realization device by reducing resolution prediction high-definition image coding unit CU dividing mode
CN103164436B (en) A kind of image search method and device
CN105976409A (en) Image compression method based on compression perception theory
CN110419214A (en) Intra prediction mode searching method and device, method for video coding and device and recording medium
CN101742305B (en) Method for estimating motions of scalable video coding and decoding system based on Markov chain model
Deng et al. Learning contextual transformer network for image inpainting
CN105872549A (en) Block search and orthogonal matching pursuit based video converting and encoding method
CN106488236B (en) The calculation method and device of absolute transformed error sum in a kind of Video coding
CN110489606A (en) A kind of grouping Hilbert coding and decoding methods
CN103561265A (en) Method for integer DST conversion base selection for video coding
CN105898305A (en) Image compression and decompression method based on JPEG-LS
CN105740428A (en) B+ tree-based high-dimensional disc indexing structure and image search method
CN107077481B (en) Information processing apparatus, information processing method, and computer-readable storage medium

Legal Events

Date Code Title Description
C06 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: 20140205