WO2009015600A1

WO2009015600A1 - The transform method and device applying for video and image process

Info

Publication number: WO2009015600A1
Application number: PCT/CN2008/071798
Authority: WO
Inventors: Lu Yu; Cixun Zhang; Zhibo Ni
Original assignee: Huawei Technologies Co., Ltd.
Priority date: 2007-07-29
Filing date: 2008-07-29
Publication date: 2009-02-05
Also published as: CN101282476A; CN101106714A; CN101282476B

Abstract

A transform method which is applied for video and image processing includes inverse discrete cosine transform method: a inversed transform preprocess step, one-dimensional inversed transform step, a inversed transform post-process step; another transform method which is applied for video and image processing includes discrete cosine transform method: positive transform preprocess step, one-dimensional positive transform step and positive transform post-process step; and this invention provides a corresponding transform process device. The simpler structure could be used to achieve the discrete cosine transform and inverse discrete cosine transform (DCT/IDCT) according to this invention and lessens the process complexity of video or image and makes higher fidelity. It approximates the ideal IDCT and DCT with much higher fidelity than the demand of associated standards and hence reduces the complication degree of the hardware. The method can be applied to various existing popular international video codec standard.

Description

Conversion method and device applied to video and image processing The application claims are filed on July 29, 2007, and December 07, 2007, respectively, and the application numbers are 200710070256.X, 200710300902.7, and the invention names are all applied. Priority of the Chinese Patent Application for "Transformation of Video and Image Processing", the entire contents of which are incorporated herein by reference. Technical field

The present invention relates to the field of electrical digital data processing technologies, and in particular, to a method and apparatus for transforming video and image processing. Background technique

In some video coding standards, such as H.261, H.263, and MPEG-1, MPEG-2, and MPEG-4 standards developed by the ISO MPEG organization, Discrete Cosine Transform is used. , DCT) removes the information redundancy of the spatial domain to achieve the purpose of compression. The ideal 8x8 discrete cosine transform and inverse discrete cosine transform are as follows:

(2m + \)k (2n + ΐ)1π

y(k, 1) = c(k)c(l) J- _ ^ ^ x(m, ) cos - cos -,

8 X 8 _m =0 n=0 16 16

1 k = 0 "i 1 = 0

Where c(A) = r- c(/) =

V2 A = other 1 2 / = Others In the ideal 8x8 discrete cosine transform, the transformation matrix multiplication coefficients are irrational numbers, and in practical implementations, these multiplication coefficient bit widths are unlikely to reach infinite length, so they are all implemented using fixed points. The method fits these irrational numbers to implement a discrete cosine transform. Similarly, in the decoding part, since the ideal 8x8 Inversed Discrete Cosine Transform (IDCT) uses the transposed matrix of the ideal 8x8 discrete cosine transform matrix, the transform matrix coefficients are also irrational, and also need Fixed point implementations to fit these irrational numbers.

In addition, in the specific implementation of the software hardware, in order to further reduce the resource overhead and improve the processing speed, the adder and the shifter are usually used instead of the multiplier function; The same fixed point number, the required adder and the number of shifts are different, which requires how to use the optimal or near optimal method to obtain the fixed point representation of these irrational numbers in the ideal 8x8 DCT/IDCT. It achieves high enough accuracy while minimizing the number of adders and shifts required.

The method of fixed-point inverse discrete cosine transform has a series of requirements on accuracy. The existing entire test is divided into random number generation, random number input test, all zero input test and adjacent DC inversion test and corresponding indicators. For example, 8x8 DCT matrix operation / _z enter the random number test, and then rounding to give 8x8 integer transform coefficient matrix F _z must preclude the floating-point precision with at least 64 bits. The IDCT operation is performed on the 8x8 transform coefficient matrix F'z, and then rounded to obtain an 8x8 matrix. At least 64 bits of floating point precision must be used. In the All Zero Test, set F[v][3⁄4] = 0, M = 0..7 and v = 0..7. A measured IDCT is performed on the 8χ8 transform coefficient matrix F[V] [M], and then rounded off. In the Near-DC Inversion Test, the corresponding 8 χ 8 matrix, the ID CT of the 8 χ 8 transform coefficient matrix F _z "tested once, and then rounded to get the corresponding 8x8 matrix 2' _Z. It can be seen that the existing fixed point conversion method for video and image processing is complicated and the precision is not high.

Embodiments of the present invention provide a transform method and apparatus applied to video and image processing, which can easily implement 8x8 discrete cosine transform/inverse discrete cosine transform (DCT/IDCT), reduce processing complexity of video or image, and have comparison High precision.

A transform method applied to video and image processing provided by an embodiment of the present invention includes an 8x8 inverse discrete cosine transform method:

Multiply a predetermined 8x8 integer matrix M with an 8x8 input data block to obtain a matrix X of 8x8;

The matrix X is generated into eight sets of first vectors in a row, and the eight sets of vectors are sequentially subjected to one-dimensional inverse transformation to obtain a matrix χ';

The matrix x' is generated into eight sets of second vectors in rows, and the eight sets of vectors are sequentially inversely transformed in one dimension to obtain a matrix X";

The elements in the matrix X" are shifted right by w bit shifting, where w is a positive integer. Another transform method applied to video and image processing provided by an embodiment of the present invention includes an 8x8 discrete cosine transform:

Shifting the elements in the 8x8 input data block y to the left by r bits to obtain the matrix y', where r is a positive integer;

The matrix y' is generated into eight sets of first vectors in a row, and the eight sets of vectors are sequentially subjected to one-dimensional forward transform to obtain a matrix y";

The matrix y" is generated into eight sets of second vectors in rows, and the eight sets of vectors are sequentially subjected to one-dimensional forward transform to obtain a matrix y";

The shifting operation of the elements in the output matrix obtained by the one-dimensional forward transform to the right by z bits yields a matrix Y, where z is a positive integer.

A transform processing apparatus for video and image processing is provided for implementing an 8x8 inverse discrete cosine transform, and the apparatus includes:

The preprocessing unit multiplies an 8x8 integer matrix M by a predetermined 8x8 input data block to obtain an 8x8 matrix X;

a first inverse transform unit, generating, by the matrix X, eight sets of first vectors, and sequentially performing one-dimensional inverse transform on the eight sets of vectors to obtain a matrix x;

a second inverse transform unit, which generates eight sets of second vectors by rows, and sequentially performs one-dimensional inverse transform on the eight sets of vectors to obtain a matrix X";

The post-processing unit performs the right shift w bit shifting of the elements in the matrix X", where w is a positive integer.

Another transform processing apparatus applied to video and image processing according to an embodiment of the present invention is used to implement an 8x8 discrete cosine transform, including:

The preprocessing unit shifts the elements in the 8x8 input data block y to the left by r bits to obtain a matrix Y, and r is a positive integer;

The positive transform unit uses the matrix Y obtained by the pre-processing to be arranged into eight sets of vectors in columns, and uses at least six different multiplication coefficients β, b, c, respectively, when performing multiplication in each set of vector processing. d, h, ί perform operations;

The post-processing unit performs a right shifting operation of the elements in the output matrix obtained by the one-dimensional forward transform, where ζ is a positive integer. The invention is applied to the inverse transform and forward transform method of video and image processing, and can well approximate the ideal 8 x 8 inverse discrete cosine transform and the ideal 8 x 8 discrete cosine transform, and the precision far exceeds the existing DCT/ The requirements of IDCT, at the same time, the complexity of hardware implementation is greatly reduced, and can be applied to a variety of existing popular international video codec standards. The inverse transform preprocessing process can be combined with the inverse quantization of video compression decoding to further reduce the complexity.

DRAWINGS

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the description of the prior art will be briefly described below. Obviously, the following drawings are the present invention. Some of the embodiments may be obtained by those of ordinary skill in the art without departing from the drawings.

1 is a flowchart of an inverse transform method applied to video and image processing according to an embodiment of the present invention; FIG. 2 is a flowchart of a forward transform method applied to video and image processing according to an embodiment of the present invention; FIG. 3 is an 8x8 inverse discrete embodiment of the present invention. Flowchart of one-dimensional inverse transformation of cosine transform; Figure 4 is a flow chart of one-dimensional forward transform of an 8x8 discrete cosine transform according to an embodiment of the present invention. detailed description

The technical solutions in the embodiments of the present invention are clearly and completely described in the following with reference to the accompanying drawings in the embodiments of the present invention. It is obvious that the described embodiments are only a part of the embodiments of the present invention, but not all of the embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments without departing from the inventive scope are the scope of the invention.

Embodiment 1

A transform method applied to video and image processing provided by this embodiment includes an 8x8 matrix M in the inverse transform preprocessing step of the 8x8 inverse discrete cosine transform:

A B C D A D C B

B E F G B G F E

C F H I C I H F

D G I J D J I G

A B C D A D C B

D G I J D J I G

C F H I C I H F

BEFGBGFE Take the contact parameter =10, k=\2, Ψ = 1.3522468075656300,

^ =0.9596162302465070 can get:

^4=1024.0000000000000000, 5=757.2582122367510000,

C=1067.0932480341200000,, =1070.9248339636200000,

=560.0000000000000000 , =789.1260989220920000 ,

G=791.9595949289330000, H=\ 112.0000000000000000,

7=1115.9928315182000000, J=l 120.0000000000000000;

Ax2 ⁿ = 5432.3763787064400000 , bx2 ¹² = 1080.5668459858300000 , cx2 ¹² = 4605.3463196628800000, χ 2 ¹² = 3077.1940310190000000, 2 2 ¹² = 2127.2189396844700000,

5135.5608143231700000. Take ^ = 1024, Β=Ί5Ί, C=1067, =1071, =560, =789, G=792, 77=1112,

2=2128/4096, i=5136/4096.

The inverse discrete cosine transform method applied to video and image processing described in this embodiment has the following detailed steps:

1. Input data 8x8 data block X, do operation I to get the inverse transform preprocessed output number

Where Λ ·][/Ί represents the (zj)th element in the matrix M;

2. Perform the first one-dimensional inverse transformation. The steps are as follows:

Arrange X into 8 groups of vectors, and take the first set of vectors ([0][0], [0][1], ···, [0][7]) as input, and perform the following operations:

Ρ10= [0][0], ρ14= [0][4], ρ\2=Χ[0][2], ρ\6=Χ[0][6],

Ρ17= [0][1]- [0][7], ρ13= [0][3], ρ\5=Χ[0][5], pll= [0][l]+ [0][ 7]; ρ20=ρ10+ρ14, ρ24=ρ10-ρ14, p22=pl2*h-pl6*t, p26=pl6*h+pl2*t, ρ27=ρ15+ρ17, p23=pll-pl3, ρ25=ρ17 -ρ15, p21=pll+pl3;

Ρ30=ρ20+ρ26, ρ34=ρ24+ρ22, ρ32=ρ24-ρ22, ρ36=ρ20-ρ26, p37=p27*c-p21*d, p33=p23*a-p25*b, ρ35=ρ23 *b+ P25 *a , P31=p21*c+p27*d;

'[0][0]=p30+p31, '[l][0]=p34+p35, '[2][0]=p32+p33, '[3][0]=p36+p37, '[ 4][0]=p36-p37, '[5][0]=p32-p33, '[6][0]=p34-p35, '[7][0]=p30-p31; where * means multiplication Operation

3. Repeat step 2, ( [1][0], [1][1],···, [1][7]), ( [2][0], [2][1],·· ·, [2][7]),

([3][0], [3][1],···, [3][7]) , ( [4][0], [4][1],···, [4][ 7]),

([5][0], [5][1],···, [5][7]) , ( [6][0], [6][1],···, [6][ 7]),

([η[ο], [ ⁷ ][ι],···, [ ⁷ ][η) sequentially perform the same one-dimensional inverse transformation to obtain the first one-dimensional inverse transform output X';

4. Perform the second one-dimensional inverse transformation. The steps are as follows:

Arrange X, by row into 8 sets of vectors, and take the first set of vectors (χ'[0][0], χ'[0][1], ···, χ'[0][7]) as input , do the following:

Ρ10= '[0][0], ρ14= '[0][4], ρ12= '[0][2], ρ16= '[0][6],

1717= '[0][1]- '[0][7], ρ13= '[0][3], ρ15= '[0][5], pll= '[0][1]+ '[ 0][7]; ρ20=ρ10+ρ14, ρ24=ρ10-ρ14, p22=pl2*h-pl6*t, p26=pl6*h+pl2*t, ρ27=ρ15+ρ17, p23=pll-pl3, 2525=ρ17-ρ15, p21=pll+pl3;

Ρ30=ρ20+ρ26, ρ34=ρ24+ρ22, ρ32=ρ24-ρ22, ρ36=ρ20-ρ26,

P37=p27*c-p21*d, p33=p23*a-p25*b, ρ35=ρ23 *b+p25 *a ,

P31=p21*c+p27*d;

χ"[0][0]=ρ30+ρ31, χ"[1][0]=ρ34+ρ35,

χ"[3][0]=ρ36+ρ37, χ"[4][0]=ρ36-ρ37, χ"[5][0]=ρ32-ρ33, χ"[6][0]=ρ34- Ρ35,

5. Repeat step 4 to set (Χ'[1][0],Χ'[1][1],···,Χ'[1][7]),(Χ'[2][0], Χ'[2][1],···,Χ'[2][7]), (χ'[3][0],χ'[3][1],···,χ'[3 ][7]) , (Χ'[4][0],Χ'[4][1],···,Χ'[4][7]),

(Χ'[5][0],Χ'[5][1],···,Χ'[5][7]) , (χ'[6][0],χ' just,...,χ '[6][7]) ,

(χ'[η[ο],χ'[ ⁷ ][ι] ··,χ'[ ⁷ ][η) performs the same one-dimensional inverse transformation to obtain the second one-dimensional inverse transformation output X";

6. The obtained output of the second one-dimensional inverse transform X" is post-processed and shifted by w bits to obtain the output X of the 8x8 inverse discrete cosine transform, where w=13, the operation is as follows: x[v][w] = x"[v][w]»13 v, we [0,63], >> denotes a right shift operation, and conversely, "represents a left shift operation."

A transform method applied to video and image processing, which includes an 8x8 matrix M in an inverse transform preprocessing step of an 8x8 inverse discrete cosine transform:

A B C D A D C B

B E F G B G F E

C F H I C I H F

D G I J D J I G

A B C D A D C B

D G I J D J I G

C F H I C I H F

B E F G B G F E Take contact parameters =10, k=U, ^ =0.9000703207408190, =

0.5918825969335950 can get:

^4=1024.0000000000000000, 5=1137.6888854164000000,

C=1730.0728308369000000, =1608.9350515170000000,

£=1264.0000000000000000 , =1922.1529595742400000,

G=1787.5659428395900000, 77=2923.0000000000000000,

7=2718.3347843854700000, J=2528.0000000000000000;

719.2371556781520000, cx2 ¹² = 3065.3690701062000000, ί 2 ¹² =2048.2141299835100000, hx2 ^l2 = 1312.0493699271300000, ί 2 ¹² = 3167.5673833811500000, take ^ =1024, β=1138, C=1730, =1609, £=1264, =1922, G=l 788, 77=2923,

2=1312/4096, i=3168/4096 The inverse discrete cosine transform method applied to video and image processing, with detailed steps: 1. The 8x8 data X to be transformed is subjected to inverse transform preprocessing to obtain the inverse transform preprocessed input.

Arrange the X rows into 8 groups of vectors, and take the first row vector ( [ο][ο], [ο][ι],···, [ο][η) as input, and perform the ^ mouth operation:

Ρ10= [0][0], ρ14= [0][4], ρ\2=Χ[0][2], ρ\6=Χ[0][6],

Ρ17= [0][1]- [0][7], ρ13= [0][3], ρ\5=Χ[0][5], pll= [0][l]+ [0][ 7]; ρ20=ρ10+ρ14, ρ24=ρ10-ρ14, p22=pl2〇h-pl6〇t,

P26=pl60h+pl20t, ρ27=ρ15+ρ17, p23=pll-pl3, ρ25=ρ17-ρ15, p21=pll+pl3;

3030=ρ20+ρ26, ρ34=ρ24+ρ22, ρ32=ρ24-ρ22, ρ36=ρ20-ρ26, p37=p27〇c-p21〇d, p33=p23〇a-p25〇b, p35=p23〇b+ P25〇a, p31=p21〇c+p27〇d;

'[0][0]=p30+p31, '[l][0]=p34+p35, '[2][0]=p32+p33, '[3][0]=p36+p37, '[ 4][0]=p36-p37, '[5][0]=p32-p33, '[6][0]=p34-p35, '[7][0]=p30-p31 where 〇 indicates Add (subtract) method, shift to achieve multiplication operation, defined as follows:

X0a=l-(x»3) + (x»7);

X0b=(x»3) - (x»7) + (((x»3) - (x»7)-(x»ll))»l)

X0h= ((x+(x»5))»2) +(x»4)

X0t= x+(x»5) - ((x+(x»5))»2)

X0d=x»2

X0c=(((x»9)-x)»2)- ((x»9)-x) 3. Repeat step 2, ( [1][0], [1][1],···, [1][7]) , ( [2][0], [2][1] ,-,Χ[2][7]),

([3][0], [3][1],···, [3][7]) , ( [4][0], [4][1],···, [4][ 7]),

([5][0], [5][1],···, [5][7]) , ( [6][0], [6][1],···, [6][ 7]),

([ ⁷ ][ο], [ ⁷ ][ι],···, [ ⁷ ][η) perform the same one-dimensional inverse transformation to get the first one-dimensional inverse transformation Output Χ';

Arrange X, by row into 8 sets of vectors, and take the first row vector (χ'[0][0], χ'[0][1],···,χ'[0][7]) as input , do the following:

Ρ10= '[0][0], ρ14= '[0][4], ρ12= '[0][2], ρ16= '[0][6],

Ρ17= '[0][1]- '[0][7], ρ13= '[0][3], ρ15= '[0][5],

Pll= '[0][l]+ '[0][7];

2020=ρ10+ρ14, ρ24=ρ10-ρ14, p22=pl2〇h-pl6〇t,

P26=pl60h+pl20t, ρ27=ρ15+ρ17, p23=pll-pl3, ρ25=ρ17-ρ15, p21=pll+pl3;

x"[0][0]=p30+p31, x"[l][0]=p34+p35, x"[2][0]=p32+p33,

x"[3][0]=p36+p37, x"[4][0]=p36-p37, x"[5][0]=p32-p33,

x"[6][0]=p34-p35, x"[7][0]=p30-p31

The definition of the operation is the same as the definition in step 2;

5. Repeat step 4 to set (Χ'[1][0],Χ'[1][1],···,Χ'[1][7]),(Χ'[2][0], Χ'[2][1] ··,Χ'[2][7]), (χ'[3][0],χ'[3][1],···,χ'[3][ 7]) , (Χ'[4][0],Χ'[4][1],···,Χ'[4][7]),

(Χ'[5][0],Χ'[5][1],···,Χ'[5][7]) , (Χ'[6][0],Χ' just,...,χ '[6][7]) ,

6. The output of the second one-dimensional transformed output X" is post-processed to the right by w bits to obtain the output X of the 8x8 inverse discrete cosine transform, where w=13, the steps are as follows:

x[v][w] = x"[v][w]»13 v,we[0,63].

A transform method applied to video and image processing, which includes an 8x8 matrix M in an inverse transform preprocessing step in an 8x8 inverse discrete cosine transform: ABCDADCB

B E F G B G F E

C F H I C I H F

D G I J D J I G

A B C D A D C B

D G I J D J I G

C F H I C I H F

B E F G B G F E take contact parameters =10, k=\2,

Ψ = 0.9000703207408190, = 0.5918825969335950 can get:

^4=1024.0000000000000000, 5=1137.6888854164000000,

C=1730.0728308369000000, =1608.9350515170000000,

£=1264.0000000000000000 , =1922.1529595742400000,

G=1787.5659428395900000, 77=2923.0000000000000000,

7=2718.3347843854700000, J=2528.0000000000000000;

719.2371556781520000, cx2 ¹² = 3065.3690701062000000, ί 2 ¹² =2048.2141299835100000, hx2 ^l2 = 1312.0493699271300000, ί 2 ¹² = 3167.5673833811500000. Take ^ =1024, β=1138, C=1730, =1609, =1264, =1922, G=1788, 77=2923, 7=2718, J=2528, a=3616/4096, b=719/4096, c=3066/4096,

The corresponding 8x8 inverse discrete cosine transform method consisting of the above { , B, C, D, E, F, G, H, I, J, a, b, c, d, 2, i } Accuracy requirements in accordance with ISO/IEC 23002-1-2006 and the following two definitions:

(a) for e [0,63], e [1,255], yielding 64 input data

N _{i ;} [x/S][x%S] =

0, other

Let 8 8 input the data matrix ^, . _; .=-^ _; .; will Ν _ί; ^οΝ'_ί; . respectively by { B, C, D, E, F, G, H, I, J, a 8x8 inverse discrete cosine transform composed of b, c, d, h, ί } Change, get two output blocks, add them to get z, all combinations in the range of values, the resulting ^ is all zero;

(b) A random number generation method defined in accordance with the ISO/IEC 23002-1-2006 standard, for random numbers in the range 1,000,000 [-1, +1], these random numbers are taken as inputs, and {^, B, The 8x8 inverse discrete cosine transform composed of C, D, E, F, G, H, I, J, a, b, c, d, h, ί } is tested to obtain omse = 0.000204, less than 0.01.

The inverse discrete cosine transform method applied to video and image processing has detailed steps:

1. The 8x8 data X to be transformed is inversely preprocessed to obtain the output of the inverse transform preprocessing: ΜΜ = ί 'just ^{XM[V][W] +} 2 ¹²¹ , ^{V = U =} °

~ {X v][u]xM[v][u] ,Other

2. Perform the first one-dimensional inverse transformation. The steps are as follows: Arrange the X rows into 8 groups of vectors, and set the first row vector ([0][0], [0][1],···, [0 ][7]) As input, perform the following operations:

Ρ10= [0][0], ρ14= [0][4], ρ12= [0][2], ρ16= [0][6],

Ρ17= [0][1]- [0][7], ρ13= [0][3], ρ15= [0][5], pll= [0][l]+ [0][7]; 2020=ρ10+ρ14, ρ24=ρ10-ρ14, p22=pl2〇h-pl6〇t,

P26=pl60h+pl20t, ρ27=ρ15+ρ17, p23=pll-pl3, ρ25=ρ17-ρ15, p21=pll+pl3;

'[0][0]=p30+p31, '[l][0]=p34+p35, '[2][0]=p32+p33, '[3][0]=p36+p37, '[ 4][0]=p36-p37, '[5][0]=p32-p33, '[6][0]=p34-p35, '[7][0]=p30-p31 where 〇 operation definition as follows:

X0a=l-(x»3) + (x»7);

X0b=(x»3) - (x»7) + (((x»3) - (x»7)-(x»ll))»l)

X0h= ((x+(x»5))»2) +(x»4)

X0t= x+(x»5) - ((x+(x»5))»2)

X0c=(((x»9)-x)»2)- ((x»9)-x);

3. Repeat step 2, ( [1][0], [1][1],···, [1][7]), ( [2][0], [2][1],·· ·, [2][7]), ( [3][0], [3][1],···, [3][7]) , ( [4][0], [4][1 ],···, [4][7]) ,

([5][0], [5][1],···, [5][7]) , ( [6][0], [6][1],···, [6][ 7]),

([η[ο], [ ⁷ ][ι],···, [ ⁷ ][η) performs the same one-dimensional inverse transformation to obtain the first one-dimensional inverse transformation output χ';

Arrange X, by row into 8 sets of vectors, and take the first row vector (X '[0] [0], X '[0] [1], ···, X '[0] [7]) as input , do the following:

Ρ10= '[0][0], ρ14= '[0][4], ρ12= '[0][2], ρ16= '[0][6],

Ρ17= '[0][1]- '[0][7], ρ13= '[0][3], ρ15= '[0][5],

Pll= '[0][l]+ '[0][7];

2020=ρ10+ρ14, ρ24=ρ10-ρ14, p22=pl2〇h-pl6〇t,

P26=pl60h+pl20t, ρ27=ρ15+ρ17, p23=pll-pl3, ρ25=ρ17-ρ15, p21=pll+pl3;

x"[0][0]=p30+p31, x"[l][0]=p34+p35, x"[2][0]=p32+p33,

x"[3][0]=p36+p37, x"[4][0]=p36-p37, x"[5][0]=p32-p33, x"[6][0]=p34- P35, x"[7][0]=p30-p31

The operation of 〇 is the same as in 2;

(χ'[5][0],χ'[5][1],···,χ'[5][7]) , (Χ'[6][0],Χ' just,...,χ '[6][7]) ,

(χ'[ ⁷ ][0],χ'[ ⁷ ][1] ··,χ'[ ⁷ ][η) performs the same one-dimensional inverse transformation to obtain the second one-dimensional inverse transformation output X";

6. The obtained output of the second one-dimensional transformed X" is post-processed and shifted right by w bits. The output x of the 8x8 inverse discrete cosine transform, where w=13, is as follows:

x[v][u] = x"[v][u]»\3 v,we [0,63].

A transform method applied to video and image processing, which includes an 8x8 matrix M in a post-transformation processing step in an 8x8 discrete cosine transform:

A B C D A D C B

B E F G B G F E

C F H I C I H F

D G I J D J I G

A B C D A D C B

D G I J D J I G

C F H I C I H F

B E F G B G F E take contact parameters =10, k=U,

Ψ = 1.1812048201875100, ^=0.8012529373378150, you can get:

^4=1024.0000000000000000, 5=866.9114640401220000,

C=1277.9984350538200000, =1225.9979498222600000,

£=733.9213735197150000, =1081.9448187241100000,

G=1037.9215601470700000, 77=1595.0000000000000000,

7=1530.1010363789400000, J=1467.8427470394300000;

Ax2 ^l4 = 18981.0000000000000000, bx2 ¹⁴ =3775.5556452332900000, cx2 ¹⁴ =16091.3148131935000000, ίχ2 ¹⁴ =10751.8728142102000000, 2 2 ¹⁴ =7104.6752652150500000, t>2 ^l4 = 17152.2033815388000000. Take ^ =1024, β=867, C=1278, =1226, £ =734, =1082, G=1038, 77=1595, 7=1530, J=1468, «=18981/16384, b=3776/16384, c=16091/16384,

The discrete cosine transform method applied to video and image processing has the following detailed steps:

1. The 8x8 data y to be transformed is subjected to positive transform preprocessing to obtain the output of the forward transform preprocessing: ''[V][M] = '[M][V] « 7, v,we[0,7];

2. Perform the first one-dimensional positive transformation. The steps are as follows:

Arrange y, arranged into 8 groups of vectors by column, and take the first column vector Cy'[0][0], _y'[l][0], .i'[7][0]f as input, and perform the following operations :

Ql0=y,[0][0]+y,[l][0], ql4=y,[l][0]+y,[6][0], ql2=y'[2][0] +y'[5][0], ql6=y'[3][0]+y'[4][0], ql7=y'[3][0]-y'[4][0] , Ql3=y'[2][0]-y'[5][0] , ql 5=y, [ 1 ] [0]-y, [6] [0] , qll =y, [0] [0 ] -y, [7] [0];

Q20=ql0+ql6 , q24=ql4+ql2 , q22=ql4-ql2 , q26=ql0-ql6 , q27=ql7*c+qll*d , q23=ql3*a+ql5*b , q25=ql5*a- Ql3*b , q21=qll*c-ql7*d;

Q30=q20+q24, q34=q20-q24, q32=q22*h+q26*t, q36=q26*h-q22*t, q37=q27+q25, q33=q21-q23, q35=q27-q25, Q31=q21+q23;

y"[0][0]=q30, y"[0][l]=q31+q31, y"[0][2]=q32, y"[0][3]=q33, y"[0 ][4]=q34, y"[0][5]=q35, y"[0][6]=q36, y"[0][7]=q31-q37;

3. Repeat step 2 to ([0][l], [l][l],.i'[7][l]f, ([0][2], [1][2],·· · [7][2]) ^Τ ,

([0][3], [1][3],··· [7][3]) ^Τ , ( [0][4], [1][4],··· [7][4 ]) ^Τ ,

([0][5], [1][5],··· [7][5]) ^Τ , ( [0][6], [1][6],··· [7][6 ]) ^Τ ,

([0][7], [1][7],··· [7][7]) ^Τ , perform the same one-dimensional forward transformation, and obtain the first one-dimensional forward transform output y";

4. Perform the second one-dimensional positive transformation. The steps are as follows:

Arrange y" into 8 groups of vectors, and take the first column vector Cy"[0][0], _y"[l][0],..._y"[7][0]f as input, The following operations:

Ql0=y"[0][0]+y"[l][0], ql4=y"[l][0]+y"[6][0],

Ql2=y"[2][0]+y"[5][0], ql6=y"[3][0]+y"[4][0], ql7=y"[3][0] -y"[4][0], ql3=y"[2][0]-y"[5][0], ql5=y"[l][0]-y"[6][0] , Qll=y"[0][0]-y"[7][0]; q20=ql0+ql6 , q24=ql4+ql2 , q22=ql4-ql2 , q26=ql0-ql6 , q27=ql7*c+ Qll*d , q23=ql3*a+ql5*b , q25=ql5*a-ql3*b , q21=qll*c-ql7*d;

y"'[0][0]=q30, y"'[0][l]=q31+q31 , y"'[0][2]=q32 , y"'[0][3]=q33 , y"'[0][4]=q34, y"'[0][5]=q35, y"'[0][6]=q36, y"'[0][7]=q31-q37;

5. Repeat step 4 to set ( '[0][1], '[1][1],... '[7][1]) ^Γ ,

( '[0][2], '[1][2],··· '[7][2]) ^Γ , ( '[0][3], '[1][3],··· '[7][3]) ^Γ ,

( '[0][4], _y"[l][4],.i"[7][4]) ^r , ( '[0][5], [1][5],·ι"[ 7][5]) ^Γ ,

Cy"[0][6],_y"[l][6],'i"[7][6]f, Cy"[0][7],_y"[l][7],'i" [7][7]f

Perform the same one-dimensional forward transformation to obtain a second one-dimensional forward transform output y",;

6. The obtained second-dimensional one-dimensionally transformed output y" is post-processed and shifted to the right by the z-bit to obtain the output Y of the 8x8 discrete cosine transform, taking z=20, as follows:

Y[v][u] = [y" v][u]xM[v][u] + (l « 19)- /(,[ ])] » 20. Example 5

A B C D A D C B

B E F G B G F E

C F H I C I H F

D G I J D J I G

A B C D A D C B

D G I J D J I G

C F H I C I H F

B E F G B G F E

Take the contact parameter =10, k=U,

Ψ = 0.9000703207408190, = 0.5918825969335950 can get: ^4=1024.0000000000000000, 5=1137.6888854164000000,

C=1730.0728308369000000, =1608.9350515170000000,

£=1264.0000000000000000 , =1922.1529595742400000,

G=1787.5659428395900000, 77=2923.0000000000000000,

7=2718.3347843854700000, J=2528.0000000000000000;

719.2371556781520000, Cx2 ¹² = 3065.3690701062000000, ί 2 ¹² =2048.2141299835100000, hx2 ^l2 = 1312.0493699271300000, ί 2 ¹² = 3167.5673833811500000. Take ^ =1024, β=1138, C=1730, =1609, =1264, =1922, G=1788, 77= 2923, 7=2718, J=2528, a=3616/4096, b=719/4096, c=3066/4096,

1. The 8x8 data y to be transformed is pre-processed to obtain the output of the positive transform preprocessing: ''[V][M] = '[M][V] « 7, v, we[0,7];

2. Arrange y' into 8 groups of vectors by column, and take the first column vector ( '[0] [0], y '[1][0],... '[7][0]f as input, and proceed as follows Operation:

Ql0=y,[0][0]+y,[l][0], ql4=y,[l][0]+y,[6][0] , ql2=y'[2][0] +y'[5][0], ql6=y'[3][0]+y'[4][0], ql7=y'[3][0]-y'[4][0] , Ql3=y'[2][0]-y'[5][0] , ql 5=y, [ 1 ] [0]-y, [6] [0] , qll =y, [0] [0 ] -y, [7] [0];

Q20=ql0+ql6 , q24=ql4+ql2 , q22=ql4-ql2 , q26=ql0-ql6 , q27=ql7〇c+qll〇d, q23=ql3〇a+ql5〇b , q25=ql5〇a- Ql3〇b , q21=qll〇c-ql7〇d;

Q30=q20+q24, q34=q20-q24, q32=q22〇h+q26〇t,

Q36=q26〇h-q22〇t, q37=q27+q25, q33=q21-q23, q35=q27-q25, q31=q q37=q27+q25, q33=q21-q23, q35=q27-q25, q31 =q21+q23;

The 〇 operation is defined as follows:

X0a=l-(x»3) + (x»7);

X0b=(x»3) - (x»7) + (((x»3) - (x»7)-(x»ll))»l)

X0h= ((x+(x»5))»2) +(x»4)

X0t= x+(x»5) - ((x+(x»5))»2)

X0d=x»2

X0c=(((x»9)-x)»2)- ((x»9)-x); 3. Repeat step 2 to ([0][l], [l][l],.i'[7][l]f, ([0][2], [1][2],·· · [7][2]) ^Τ , ( [0][3], [1][3],··· [7][3]) ^Τ , ( [0][4], [1][4 ],··· [7][4]) ^Τ ,

([0][5], [1][5],··· [7][5]) ^Τ , ( [0][6], [1][6],··· [7][6 ]) ^Τ ,

([0][7], [1][7],··· [7][7]) ^Τ Perform the same one-dimensional forward transformation to obtain the first one-dimensional forward transform output y";

Ql0=y"[0][0]+y"[l][0], ql4=y"[l][0]+y"[6][0], ql2=y"[2][0] +y"[5][0] , ql6=y"[3][0]+y"[4][0], ql7=y"[3][0]-y"[4][0], Ql3=y"[2][0]-y"[5][0] , ql5=y"[l][0]-y"[6][0], qll=y"[0][0] -y"[7][0];

Q20=ql0+ql6, q24=ql4+ql2, q22=ql4-ql2, q26=ql0-ql6, q27=ql7〇c+qll〇d, q23=ql3〇a+ql5〇b, q25=ql5〇a- Ql3〇b, q21=qll〇c-ql7〇d;

Q30=q20+q24, q34=q20-q24, q32=q22〇h+q26〇t,

Q36=q26〇h-q22〇t, q37=q27+q25, q33=q21-q23, q35=q27-q25, q31=q21+q23;

y"'[0][0]=q30, y"'[0][l]=q31+q31 , y"'[0][2]=q32 , y"'[0][3]=q33 , y"'[0][4]=q34, y"'[0][5]=q35, y"'[0][6]=q36, y"'[0][7]=q31-q37 The same as in step 2;

5. Repeat step 4 to set ( '[0][1], '[1][1],... '[7][1]) ^Γ ,

( '[0][2], _y"[l][2],.i"[7][2]f , ( '[0][3], '[1][3],··· ' [7][3]) ^Γ ,

( '[0][4], _y"[l][4],.i"[7][4]) ^r , ( '[0][5], [1][5],·ι"[ 7][5]) ^Γ ,

Cy"[0][6],_y"[l][6],'i"[7][6]f, ( '[0][7],_y"[l][7],.i" [7][7]f performs the same one-dimensional forward transformation to obtain a second one-dimensional forward transform output y",;

6. The obtained second-dimensional one-dimensionally transformed output y" is post-processed and shifted to the right by the z-bit to obtain the output Y of the 8x8 inverse discrete cosine transform, taking z=20, as follows:

Y[v] [u] = [y ^m [v] [u]xM[v] [u] + (1 «19)- f(y ^m [v] [u])] » 20. Embodiment 6

A B C D A D C B

B E F G B G F E

C F H I C I H F

D G I J D J I G

A B C D A D C B

D G I J D J I G

C F H I C I H F

B E F G B G F E

Take the contact parameter =10, k=\2,

C=1730.0728308369000000, =1608.9350515170000000,

£=1264.0000000000000000 , =1922.1529595742400000,

G=1787.5659428395900000, 77=2923.0000000000000000,

7=2718.3347843854700000, J=2528.0000000000000000;

Αχ2 ¹² = 3615.8493569450400000, bx2 ¹² = 719.2371556781520000, cx2 ¹² = 3065.3690701062000000, ί 2 ¹² =2048.2141299835100000 , hx2 ^l2 = 1312.0493699271300000, ί 2 ¹² = 3167.5673833811500000. Take ^ =1024, β=1138, C=1730, =1609, =1264 , =1922, G=1788, 77=2923, 7=2718, J=2528, a=3616/4096, b=719/4096, c=3066/4096,

Referring to Figure 1, the detailed steps of the inverse discrete cosine transform method applied to video and image processing are as follows:

1. Input data 8x8 data block X, dynamic range is [-2 ⁸⁺³ , 2 ⁸⁺³ - 1], that is 12 bits wide, X, input, inverse transform preprocessing:

_Υ[ \v ₁ \ _Γ \u _π [v][ _M ]xM just + 2 ¹² - ¹ , v = u = 0

X \ = <

\X v][u]xM[v][u] ,Other Obtaining the inverse transform preprocessed output data X, X has a dynamic range of [-2 ^{β + 14} , 2 ^{β + 14} _1 ] , that is, 23 bits wide;

2. Arrange X by row into 8 sets of vectors:

([0][0], [0][1],···, [0][7]) , ( [1][0], [1][1],···, [1][ 7]),

([2][0], [2][1],···, [2][7]) , ( [3][0], [3][1],···, [3][ 7]),

([4][0], [4][1],···, [4][7]) , ( [5][0], [5][1],···, [5][ 7]),

([6][0], [6][1],···, [6][7]) , ( [7][0], [7][1],···, [7][ 7]);

Each set of vectors is subjected to one-dimensional inverse transform, in which each set of vectors is inversely transformed as Χ Χ ' in Fig. 3, and Χο~χ ₇ in Fig. 3 is the result of one-dimensional inverse transform of each group:

x' ₀ =(x'[0][0],x'[0][l],-,x'[0][7]),x' ₁ =(x'[l][0],x '[l][l],-, '[l][7]),

χ' ₂ =(χ'[2][0],χ'[2][1] ··,χ'[2][7]), x' ₃ = (x'[3][0], χ '[3][1],···, χ'[3][7]),

χ' ₄ =(χ'[4][0],χ'[4][1] ··,χ'[4][7]), x' ₅ = (Χ'[5][0],Χ '[5][1] ··,Χ'[5][7]) , χ' ₆ =(χ'[6][0],χ'[6][1],···,χ'[ 6][7]), x' ₇ =(Χ'[7][0],Χ'[7][1],···,Χ'[7][7]).

The ® operations in Figure 3 are implemented using the following combination of addition, subtraction, and shift operations: x<8>a=l-(x»3) + (x»7);

x<Sb=(x»3) - (x»7) + (((x»3) - (x»7)-(x»ll))»l)

X®h= ((x+(x»5))»2) +(x»4)

x<3⁄4t= x+(x»5) - ((x+(x»5))»2)

X®d=x»2

X®c=(((x»9)-x)»2)- ((x»9)-x)

The output of the one-dimensional inverse transform x,,

The dynamic range of X, and intermediate variables does not exceed [-2 ⁸⁺¹⁷ , 2 ⁸⁺¹⁷ - 1];

3. The stored Xs are arranged into 8 groups of vectors according to the columns, and each set of vectors is respectively subjected to the second one-dimensional inverse transformation as Xo~X ₇ in Fig. 3, x in Fig. 3. ~x ₇ is the result of the second one-dimensional inverse transformation of each set of vectors, and the second one-dimensional inverse transform output X" is obtained; the input of the second one-dimensional inverse transform x, the output X" and the dynamics of the intermediate variables The range does not exceed [-2 ⁸⁺¹⁷ , 2 ⁸⁺¹⁷ - 1];

4. After X is inverse transformed, it is shifted to the right by w, where w=13:

x[v][u] = x"[v][u]»\3 v,we [0,63];

The output X of the 8x8 inverse discrete cosine transform is obtained, and the dynamic range of X is [-2 ⁸ , 2 ⁸ -1]. Example 7

A transform method applied to video and image processing, which includes an arrangement of 8x8 matrices M in a forward transform post-processing step in an 8x8 discrete cosine transform:

A B C D A D C B

B E F G B G F E

C F H I C I H F

D G I J D J I G

A B C D A D C B

D G I J D J I G

C F H I C I H F

B E F G B G F E take contact parameters =10, k=\2,

C=1730.0728308369000000, =1608.9350515170000000,

£=1264.0000000000000000 , =1922.1529595742400000,

G=1787.5659428395900000, 77=2923.0000000000000000,

7=2718.3347843854700000, J=2528.0000000000000000;

719.2371556781520000, cx2 ¹² = 3065.3690701062000000, ί 2 ¹² =2048.2141299835100000, Hx2 ^l2 = 1312.0493699271300000, ί 2 ¹² = 3167.5673833811500000. Take ^ = 1024, β=1138, C=1730, =1609, £=1264, =1922, G=1788, 77=2923, 7=2718, J=2528, a=3616/4096, b=719/4096, c=3066/4096,

Referring to Figure 2, the detailed steps of the discrete cosine transform method applied to video and image processing are as follows:

1. Input data 8x8 data block y The dynamic range is [-2 ⁸ , 2 ⁸ -1], that is, 9 bits wide, input y, and perform positive transform preprocessing: M[M] = M][V]<< 7, V, UG [0, 7], the dynamic range of the output y, y, obtained by the positive transform preprocessing is [-2 ⁸⁺⁷ , 2 ⁸⁺⁷ -1], that is, the 16-bit width;

2. Arrange y' by column into 8 groups of vectors, ( [0][0], [1][0],... [7][0]f ,

([0][l], [l][l],'1'[7][l]f , ( [0][2], [1][2],··· [7][2] ) ^Τ ,

([0][3], [1][3],··· [7][3]) ^Τ , ( [0][4], [1][4],··· [7][4 ]) ^Τ ,

([0][5], [1][5],··· [7][5]) ^Τ , ( [0][6], [1][6],··· [7][6 ]) ^Τ ,

([0][7], [1][7],··· [7][7]) ^Τ ,

Each set of vectors is used as a one-dimensional forward transform by the input 一ο~χ ₇ input one-dimensional forward transform device in Fig. 4, and the result of each set of one-dimensional forward transform is ^) & in Fig. 4:

( '[0][0], '[1][0],··· '[7][0]) ^Γ , ( '[0][1], '[1][1],·ι" [7][1]) ^Γ ,

( '[0][2], _y"[l][2],.i"[7][2]f, ( '[0][3], '[1][3],··· ' [7][3]) ^Γ ,

( '[0][4], _y"[l][4],.i"[7][4]) ^r , ( '[0][5], [1][5],·ι"[ 7][5]) ^Γ ,

Cy"[0][6],_y"[l][6],.i"[7][6]f, Cy"[0][7],_y"[l][7],'i" [7][7]f ;

The 0 operations in Figure 4 are implemented using the following combination of addition, subtraction, and shift operations: x®a=l-(x»3) + (x»7);

X®b=(x»3) - (x»7) + (((x»3) - (x»7)-(x»ll))»l);

x<3⁄4h= ((x+(x»5))»2) +(x»4);

X®t= x+(x»5) - ((x+(x»5))»2);

x<8>d=x»2;

x<8)c=(((x»9)-x)»2)- ((x»9)-x);

3. Arrange the stored y" into 8 groups of vectors, and each group of vectors is again subjected to the second one-dimensional forward transformation as x ₀ ~ x ₇ in Fig. 4, in Fig. 4. & is the vector of each group The second one-dimensional positive The result of the transformation yields the output y" of the second one-dimensional forward transform; the output of the second one-dimensional forward transform y", y", and the dynamic range of the intermediate variable does not exceed [-2 ⁸⁺¹³ , 2 ^{8 +13} - 1];

4. The output y" obtained by the second one-dimensional forward transformation is subjected to positive transformation and post-processing, that is, the z-bit is shifted right, and the output data Y of the 8x8 inverse discrete cosine transform is obtained, and z=20: that is, Y[v ][u] = [ -"'[v][w]x [v][w] + (1 « 19) -/( -'"[v][w])] » 20 ; Y Dynamic range is

[― , ]. The 8x8 inverse discrete cosine transform method corresponding to {^4, B, C, D, E, F, G, H, I, J, a, b, c, d, h, t } in the embodiment of the present invention conforms to ISO /IEC 23002-1-2006 and the accuracy requirements defined below:

(a) Linear test: For e [0,63], _/· e [1,255], generate 64 input data

N, [x/8][x%8] = |^' ^ xe[0,63], let 8 8 input data matrix ^,. _; .=-^ _; .; will Ν _{ί ;} ^οΝ'_ί; Performing an 8x8 inverse discrete cosine transform consisting of { AB, C, D, E, F, G, H, I, J, a, b, c, d, h, ί }, respectively, to obtain two output blocks, Add them to each other, and all combinations within the range of values should be all zeros;

(b) A random number generation method defined according to the ISO/IEC 23002-1-2006 standard, which generates random numbers in the range of Q groups [-L, +H], and takes these random numbers as inputs, for {^, B, The 8x8 inverse discrete cosine transform composed of C, D, E, F, G, H, I, J, a, b, c, d, h, ί } is tested to obtain the mean square error omse; where L<10, H<10, Q>=10000, and the resulting mean square error omse must be less than 0.01.

The invention is an inverse transform and forward transform method applied to video and image processing, and can well approximate an ideal 8x8 inverse discrete cosine transform and an ideal 8x8 discrete cosine transform, and the precision far exceeds ISO/IEC 23002-1-2006. The requirements, while the hardware implementation is greatly reduced in complexity, and can be applied to a variety of international video codec standards. The inverse transform preprocessing process can be combined with the inverse quantization of video compression decoding to further reduce the complexity.

Obviously, those skilled in the art should understand that the above steps of the present invention can be implemented by a general-purpose computing device, which can be concentrated on a single computing device or distributed over a network of multiple computing devices. Alternatively, they can be executed by a computing device The program code is implemented so that they can be stored in a storage device by a computing device, or they can be fabricated into individual integrated circuit modules, or the steps can be made into a single integrated circuit module. Thus, the invention is not limited to any specific combination of hardware and software.

The above embodiments are intended to illustrate and explain the principles of the invention. It is to be understood that the specific embodiments of the present invention are not limited thereto. Various changes and modifications may be made without departing from the spirit and scope of the invention. Accordingly, the scope of the invention is defined by the claims.

Claims

Rights request

A transform method applied to video and image processing, characterized in that it comprises an 8x8 inverse discrete cosine transform method:

The matrix x' is generated into eight sets of second vectors in a row, and the eight sets of vectors are sequentially subjected to one-dimensional inverse transformation to obtain a matrix X";

The elements in the matrix X" are shifted right by w bit shifting, where w is a positive integer.

The transform method applied to video and image processing according to claim 1, wherein said 8x8 matrix M is:

A B C D A D C B

B E F G B G F E

C F H I C I H F

D G I J D J I G

A B C D A D C B

D G I J D J I G

C F H I C I H F

B E F G B G F E

Wherein A, B, C, D, E, F, G, H, I, J are predetermined integer elements.

The method for transforming video and image processing according to claim 1, wherein the one-dimensional inverse transform in the 8x8 inverse discrete cosine transform method specifically includes:

Pl0= 0, pl4= 4, p\2=X2, ρ16= 6, ρ\Ί=Χ\-ΧΊ, ρ13= 3, ρ15= 5, ρ20=ρ10+ 14, ρ24=ρ10- 14, p22=pl2* H-pl6*t, p26=pl6*h+pl2*t, ρ27=ρ15+ 17, p23=pll-pl3, ρ25=ρ17-ρ15, p21=pll+pl3;

Ρ30=ρ20+ρ26, ρ34=ρ24+ρ22, ρ32=ρ24-ρ22, ρ36=ρ20-ρ26,

P37=p27*c-p21 *d, p33=p23*a-p25*b, p35=p23*b+p25*a,

P31=p21 *c+p27*d;

X0=p30+p31, xl=p34+p35, x2=p32+p33, x3=p36+p37, x4=p36-p37, X5=p32-p33, x6=p34-p35, x7=p30-p31;

XO, XI, X2, X3, X4, X5, X6, X7 are the input data of the one-dimensional inverse transformation step, xO, xl, x2, x3, x4, x5, x6, x7 are the output data of the one-dimensional inverse transformation step , plO ~pl7, p20 ~ p27, p30~p37 are intermediate variables of the one-dimensional inverse transformation step, and a, b, c, d, h, ί are constants.

4. The transform method applied to video and image processing according to claim 3, wherein the * operation is a multiplication operation, and the multiplication operation is implemented by a combination of one or more of addition, subtraction, and right shift. .

The method for transforming video and image processing according to claim 2 or 3, wherein when the dynamic range of the 8x8 input data block is [-2 ^β+3 , 2 ^β+3 _ΐ] The dynamic range of the matrix X is [-2 ^β+14 , 2 ^β+14 - 1]; the dynamic range of the matrix X, the matrix X′ and the matrix X′ does not exceed [-2 ^β+17 , 2 ^β+17 -1]; The w=13, the dynamic range of the output data after the shift processing is [-2 ^β , 2 ^β _1], which is a positive integer.

6. The transform method applied to video and image processing according to claim 2, wherein said multiplication coefficients σ, b, c, d, h, ί and parameters, B, C, D, E, F The relationship between G, H, I, and J is as follows:

A =

, ' t― 2;—— , ― υ

2 c a

^ 2

f = cos(3^-/16) , f = sin(3^-/16) , k, S, ψ, ø are the contact factors.

7. A transform method applied to video and image processing, characterized in that it comprises an 8x8 discrete cosine transform:

The matrix y' is generated into eight sets of first vectors in rows, and the eight sets of vectors are sequentially subjected to one-dimensional forward transform to obtain a matrix y"; The matrix y" generates 8 sets of second vectors in rows, and sequentially performs one-dimensional forward transformation on the 8 sets of vectors to obtain a matrix y";

The shifting operation of the output matrix y" obtained by the one-dimensional forward transform, the right element is shifted by z bits to obtain a matrix Y, where z is a positive integer.

The method for transforming video and image processing according to claim 7, wherein the one-dimensional forward transform in the 8x8 discrete cosine transform method specifically comprises:

Ql0=y0+yl, ql4=yl+y6, ql2=y2+yx5, ql6=y3+y4, ql7=y3-y4, ql3=y2-y5, ql5=yl-y6, ql l=y0-y7;

Q20=ql0+ql6, q24=ql4+ql2, q22=ql4-ql2, q26=ql0-ql6,

Q27=ql7*c+ql l *d, q23=ql3*a+ql5*b, q25=ql5*a-ql3 *b,

Q21=ql l *c-ql7*d;

Y0=q30, Yl=q31+q31, Y2=q32, Y3=q33, Y4=q34, Y5=q35, Y6=q36, Y7=q31-q37,

Where yO, yl, y2, y3, y4, y5, y6, yl are the input data of one-dimensional forward transformation, YO, Yl, Y2, Y3, Υ4, Υ5, Υ6, Υ7 are the output data of one-dimensional forward transformation, qlO ~ ql7, q20 ~ q27, q30 ~ q37 are intermediate variables of one-dimensional positive transformation, and a, b, c, d, h, ί are constants.

The method for transforming video and image processing according to claim 7, wherein the output matrix y" obtained by the one-dimensional forward transform is shifted to the right by z bits. The bit operation to get the matrix Y specifically includes:

Input the data obtained by the forward transform processing 8x8 data block y", and 8x8 matrix M do operation II to obtain the matrix Y, the operation II is:

Y[ym = [Y ^f [yWxM[y][x] + (l « \9) - f(Y y][x])] » z xe [0, 7] ; where the function m) is:

Where "represents left shift," means right shift.

The transform method applied to video and image processing according to claim 8 or 9, wherein when the dynamic range of the 8x8 input data block y is [-2 ^β , 2 ^β - 1], Place The dynamic range of the matrix y' is [-2 ^β+7 , 2 ^β+7 _1] , and the matrix y', the matrix y" and the matrix y" do not exceed [-2 ^β+13 , 2 ^β+13 - 1]; the shift parameter, the shift parameter z=20, is a positive integer.

11. The method for transforming video and image processing according to claim 7, wherein the * operation in the one-dimensional forward transform step is a multiplication operation, and the multiplication operation is performed in addition, subtraction, and right shift. One or more combinations are implemented.

12. A transform processing apparatus for video and image processing, for implementing an 8x8 inverse discrete cosine transform, characterized in that the apparatus comprises:

The post-processing unit performs the right shift of the elements in the matrix X" by w bit shifting, where w is a positive integer.

The transform processing device according to claim 12, wherein the 8x8 matrix M is:

A B C D A D C B

B E F G B G F E

C F H I C I H F

D G I J D J I G

A B C D A D C B

D G I J D J I G

C F H I C I H F

B E F G B G F E where A, B, C, D, E, F, G, H, I, J are predetermined integer elements.

The transform processing apparatus according to claim 12, wherein the inverse transform unit performs one-dimensional inverse transform according to the following steps:

Pl0= 0, pl4= 4, p\2=X2, ρ16= 6, ρ\Ί=Χ\-ΧΊ, ρ13= 3, ρ15= 5, p\\=X\+X7;

P20=pl0+ l4, p24=pl0- l4, p22=pl2*h-pl6*t, p26=pl6*h+pl2*t, p27=pl5+ l7, p23=pll-pl3, p25=pl7-pl5, p21= Pll+pl3;

P30=p20+p26, p34=p24+p22, p32=p24-p22, p36=p20-p26,

P37=p27*c-p21*d, p33=p23*a-p25*b, p35=p23*b+p25*a,

P31=p21*c+p27*d;

Where X0, XI, X2, X3, X4, X5, X6, X7 are the input data of the one-dimensional inverse transformation step, χθ, xl, x2, x3, x4, x5, x6, x7 are the output data of the one-dimensional inverse transformation step , plO ~pl7, p20 ~p27, p30 ~p37 are intermediate variables of the one-dimensional inverse transformation step, and a, b, c, d, h, t are constants.

15. A transform processing apparatus for video and image processing, for implementing an 8x8 discrete cosine transform, characterized in that:

The post-processing unit performs a right-shifting shift operation on the elements in the output matrix obtained by the one-dimensional forward transform, where ζ is a positive integer.

The conversion processing device according to claim 15, wherein the positive conversion unit performs a one-dimensional forward transformation in accordance with the following steps:

Ql0=y0+yl, ql4=yl+y6, ql2=y2+yx5, ql6=y3+y4, ql7=y3-y4, ql3=y2-y5, ql5=yl-y6, qll=y0-y7;

Q20=ql0+ql6, q24=ql4+ql2, q22=ql4-ql2, q26=ql0-ql6,

Q27=ql7*c+qll*d, q23=ql3*a+ql5*b, q25=ql5*a-ql3*b,

Q21=qll*c-ql7*d;

Y0=q30, Yl=q31+q31, Y2=q32, Y3=q33, Y4=q34, Y5=q35, Y6=q36,

Y7=q31-q37,

The conversion processing device according to claim 15, wherein the post processing unit performs processing according to the following steps:

Input the data obtained by the forward transformation processing 8x8 data block Υ, and the 8x8 matrix Μ operation II to obtain the output data of the forward transformation post-processing step Υ, the operation II is:

«Represents left shift," means right shift.