JP4451759B2 - Lossless video encoding device, lossless video encoding method, lossless video decoding device, lossless video decoding method, lossless video encoding program, lossless video decoding program, and recording medium for those programs - Google Patents

Description

  The present invention relates to high-efficiency video coding, and more particularly to lossless video coding technology and lossless video decoding technology for realizing lossless video coding having high affinity with the H.264 standard.

  Conventional lossless video encoding methods include the Motion JPEG2000 standard (see Non-Patent Literature 1 and Non-Patent Literature 2) and the JPEG-LS (see Non-Patent Literature 3 and Non-Patent Literature 4). Since these are still image bases, they are encoded in a closed frame.

These conventional methods have the following problems.
• Motion JPEG2000 can be transmitted in stages, but because it performs intra-frame coding, it cannot perform high-efficiency coding using video-specific inter-frame correlation.
・ JPEG-LS is more efficient than JPEG2000, but it is also a closed encoding within a frame, and there is a limit to efficiency, and stepwise transmission is not possible.

  Further, as a conventional lossless video encoding method, there is the second edition “Fidelity Range Extension (FRExt)” of the H.264 standard (see Non-Patent Document 5). In this method, the intraframe / interframe prediction residual signal is transmitted as it is (without orthogonal transformation / quantization). This has a high affinity with the first edition of the H.264 standard and is easy to process, but there is a problem that the correlation still remaining in the prediction residual signal cannot be removed.

In this way, no lossless video coding method has been proposed that has improved efficiency by performing inter-frame prediction and has high compression efficiency while maintaining compatibility with the H.264 standard.
ISO / IEC 15444-3: 2002 Information technology −JPEG2000 image coding system−Part 3: Motion JPEG2000 "Motion-JPEG2000", http: //e-words.jp/w/Motion-JPEG2000.html ISO / IEC 14495-1: 1999 Information technology −Lossless and near-lossless coding of continuous tone still images Ueno et al., “Multi-level lossless coding standard and adoption of our technology”, http://www.mitsubishielectric.co.jp/giho/9809/9809115.pdf Gary J. Sullivan, Pankaj Topiwala, and Ajay Luthra, “The H.264 / AVC Advanced Video Coding Standard: Overview and Introduction to the Fidelity Range Extensions”, Presented at the SPIE Conference on Applications of Digital Image Processing XXVII Special Session on Advances in the New Emerging Standard: H.264 / AVC, August, 2004

  In view of the above-described problems, the present invention aims to realize lossless video coding having high affinity with the H.264 standard.

[Definition of variables]
A small block of the original signal is represented by a 4 × 4 matrix U. In the H.264 standard, the block is predicted within a frame or between frames. This signal is also represented by a 4 × 4 matrix Y. And the prediction residual signal (4 × 4 matrix R) is the difference between them R = U−Y (1)
And All of these are integers.

  Each element of the residual signal is

のように表記することとする。

It shall be expressed as

[First invention element ]
In the H.264 standard, R is orthogonally transformed and then coefficients are quantized, but the lossless video coding method according to the present invention omits orthogonal transformation and quantization, and performs linear prediction instead. .

  Here, linear prediction is performed on each element of R to obtain a residual conversion signal. This is a 4 × 4 matrix P.

線形予測においては，一次元予測を水平垂直に繰り返す方法（アルゴリズム１）あるいは二次元予測を行う方法（アルゴリズム２）を用いる。

In linear prediction, a method of repeating one-dimensional prediction horizontally and vertically (algorithm 1) or a method of performing two-dimensional prediction (algorithm 2) is used.

[Algorithm 1]
1. (* Horizontal direction *)
2. for j ← 0 to 3
3. for i ← 0 to 3
4. p _{i, j} ← r _{i, j} -f (α _h r _{i-1, j} )
5. (* Vertical direction *)
6. for i ← 0 to 3
7. for j ← 3 downto 0
8. p _{i, j} ← p _{i, j} -f (α _v p _{i, j-1} )
[Algorithm 2]
1. for j ← 0 to 3
2. for i ← 0 to 3
3. p _{i, j} ← r _{i, j} -f (α _h r _{i-1, j} + α _v r _{i, j-1} -α _h α _v r _{i-1, j-1} )
Where r ₋₁ , _n = r _{n, −1} = 0 for any integer n
Α _h is a horizontal prediction coefficient, α _v is a vertical prediction coefficient, and a predetermined value (for example, α _h = α _v = 0.75) is used. F (x) is a function that rounds off the decimal part of x to make an integer. That is,
f (x) = [maximum integer less than (x + 0.5)] (4)
The P thus obtained is referred to as a “residual conversion signal”. This is encoded as a quantized coefficient (again 4 × 4 size) that has undergone orthogonal transformation and quantization in normal H.264.

For example, the following are used as hardware resources and operation means for performing the above linear prediction.
Integer i storage device, integer j storage device, integer r _ij storage device, integer k storage device, 16 element integer array p storage device, 16 element integer array r storage device, floating point number α _h storage device, floating point Number α _v storage device, floating point number f storage device, rounding means / multiplication means, addition means, subtraction means / integer comparison means.

The following are used as auxiliary means.
P _ij setting means (i, j, x are obtained from integer i storage device, integer j storage device and integer x storage device, i * 4 + j is stored in integer k storage device, k of integer array p storage device Means to store x in the address)
• p _ij-1 acquisition means (obtains j from integer j storage device, subtracts 1 by subtraction means, records this value in integer j storage device, executes p _ij acquisition means, and again from integer j storage device j is obtained, 1 is added by the adding means, and this value is recorded in the integer j storage device)
R _ij acquisition means (acquires i and j from integer i storage device and integer j storage device, respectively, returns 0 if i is negative or j is negative, otherwise i * 4 + j in integer k storage device And means for storing the value of address k in the integer array r storage device in the integer r _ij storage device)
R _ij-1 acquisition means (acquires j from the integer j storage device, subtracts 1 by the subtraction means, records this value in the integer j storage device, executes r _ij acquisition means, and again from the integer j storage device j is obtained, 1 is added by the adding means, and this value is recorded in the integer j storage device)
R _i-1j acquisition means (acquires i from integer i storage device, subtracts 1 by subtraction means, records this value in integer i storage device, executes r _ij acquisition means, and again from integer i storage device i means to obtain i, add 1 by adding means, and record this value in integer i storage device)
R _i-1j-1 acquisition means (acquires i from the integer i storage device, subtracts 1 by the subtraction means, records this value in the integer i storage device, executes the r _ij-1 acquisition means, and re-integers means for obtaining i from the i storage device, adding 1 by the adding means, and recording this value in the integer i storage device).

Algorithm 1 is described as the following operations on hardware resources.
[1] The value 0 is stored in the integer j storage device.
[2] The value 0 is stored in the integer i storage device.
[3] (Flag A)
[4] Obtain α _h from the α _h storage means, execute r _i-1j acquisition means, obtain the product of r _{i−1, j} value obtained from the integer r _ij storage device and α _h by the multiplication means, The product is converted to an integer by means of rounding off after the decimal point, r _ij obtaining means is executed, the difference between the r _{i, j} value obtained from the r _ij storage device and the product is obtained by the subtraction means, and stored in the integer x storage device.
[5] Execute the p _ij setting means.
[6] i is obtained from the integer i storage device, 1 is added to it by the adding means, and the result is stored in the integer i storage device.
[7] The magnitude of i and 3 is compared by an integer comparison means, and if i does not exceed 3, the process is transferred to the address of flag A.
[8] If i exceeds 3, 0 is stored in the integer i storage device, j is acquired from the integer j storage device, 1 is added thereto by the adding means, and the result is stored in the integer j storage device To do.
[9] Compare the size of j with 3, and if j does not exceed 3, the process moves to the address of flag A.
[10] The value 0 is stored in the integer i storage device.
[11] The value 3 is stored in the integer j storage device.
[12] (Flag B)
[13] Obtain α _v from the α _v storage means, execute p _ij-1 acquisition means, determine the product of pi _{, j-1} value obtained from the p _ij storage device and α _v by the multiplication means, The product is converted to an integer by means of rounding off the decimal point, p _ij obtaining means is executed, the difference between the p _{i, j} value obtained from the p _ij storage device and the product is obtained by the subtraction means, and stored in the integer x storage device.
[14] The p _ij setting means is executed.
[15] Obtain j from the integer j storage device, subtract 1 from it, and store it in the integer j storage device.
[16] Compare the magnitude of j and 0 by integer comparison means, and if j is not less than 0, move the processing to the address of flag B.
[17] If j is less than 0, 3 is stored in the integer j storage device, i is obtained from the integer i storage device, 1 is added thereto by the adding means, and the result is stored in the integer i storage device To do.
[18] Compare the size of i with 3, and if i does not exceed 3, move the process to the address of flag B.
[19] If i exceeds 3, the process is terminated.

Algorithm 2 is described as the following operations on hardware resources.
[1] The value 0 is stored in the integer i storage device.
[2] The value 0 is stored in the integer j storage device.
[3] (Flag A)
[4] Get the alpha _h from alpha _h memory means, obtains the alpha _v from alpha _v storage means.
[5] The r _ij-1 acquisition means is executed, the product of r _{i, j-1} value obtained from the r _ij storage device and α _v is obtained by the multiplication means, and stored in the floating point number f storage device.
[6] The r _i-1j acquisition means is executed, the product of the r _{i−1, j} value obtained from the r _ij storage device and α _h is obtained by the multiplication means, and added to the floating point number f storage device by the addition means. .
[7] _ri-1j-1 acquisition means is executed, the product of ri _{-1, j-1} value obtained from the r _ij storage device and α _h is obtained by the multiplication means, and the product and α _v product Is obtained by the multiplication means, and is subtracted from the floating point number f storage device by the subtraction means.
[8] The value of the floating-point number f storage device is converted into an integer by rounding means after the decimal point and stored in the integer x storage device.
[9] The r _ij obtaining means is executed, the difference between the r _{i, j} value obtained from the r _ij storage device and x is obtained by the subtraction means, and stored in the integer x storage device.
[10] The p _ij setting means is executed.
[11] i is obtained from the integer i storage device, 1 is added to it by the adding means, and the result is stored in the integer i storage device.
[12] The size of i and 3 are compared by the integer comparison means, and if i does not exceed 3, the processing is moved to the address of flag A.
[13] If i exceeds 3, 0 is stored in the integer i storage device, j is obtained from the integer j storage device, 1 is added thereto by the adding means, and the result is stored in the integer j storage device To do.
[14] Compare the size of j with 3, and if j does not exceed 3, the process moves to the address of flag A.
[15] If j exceeds 3, the process is terminated.

[Second invention element ]
The statistical properties of the prediction residual signal vary greatly depending on the image itself, but the properties of the prediction residual are considered to change depending on the prediction method, for example, when performing inter-frame prediction and intra-frame prediction. Therefore, if the block is an intra-frame prediction, the horizontal prediction coefficient and the vertical prediction coefficient are set to α _h = α _h ^intra (5)
α _v = α _v ^intra (6)
In the case of inter-frame prediction, the horizontal prediction coefficient and the vertical prediction coefficient are expressed as α _h = α _h ^inter (7)
α _v = α _v ^inter (8)
Assuming that algorithm 1 or algorithm 2 is predicted.

The value of the prediction coefficient is, for example, α _h ^intra = α _v ^intra = 0.8 (9)
α _h ^inter = α _v ^inter = 0.7 (10)
It is determined as follows.

[Third invention element ]
Experience has shown that prediction error signals have almost similar properties between adjacent blocks. Therefore, it can be expected that the prediction coefficient optimum for the neighboring block may be predicted for the block being processed.

  A prediction coefficient that minimizes the entropy of the residual transform signal P obtained by performing linear prediction on the prediction residual signal R in a certain block adjacent to the block being encoded is:

により与えられる。ただし，Ｐはアルゴリズム１あるいはアルゴリズム２により求まる残差変換信号であり，Ｈ（）はエントロピ関数である。

Given by. However, P is a residual conversion signal obtained by Algorithm 1 or Algorithm 2, and H () is an entropy function.

H (p) = − Σ _n h _n log ₂ h _n
Here, h _n is the probability that each element of P takes the value n.

In practice, since it is not easy to obtain the prediction coefficients α _h and α _v that satisfy the above, it is realistic to obtain a prediction coefficient that minimizes the power of the residual transformation signal P. This coefficient is given by the correlation coefficient σ.

The horizontal correlation coefficient σ _{h in} neighboring blocks is

垂直方向の相関係数σ_v は，

The correlation coefficient σ _{v in the} vertical direction is

により，それぞれ解析的に得られる。線形予測の範疇で残差電力を最小とする予測係数は上記で与えられるが，符号化中の画素値は復号側では知り得ないため，図１に示すような，着目ブロック（＊）に隣接した符号化済みブロックであるＮＷ（左上隣接ブロック），Ｎ（上隣接ブロック），ＮＥ（右上隣接ブロック），Ｗ（左隣接ブロック）における相関係数（符号化側・復号側ともに，可逆符号化のため知り得る情報である）を用いる。

Are obtained analytically. The prediction coefficient that minimizes the residual power in the category of linear prediction is given above, but since the pixel value being encoded cannot be known on the decoding side, it is adjacent to the block of interest (*) as shown in FIG. Correlation coefficients in the encoded blocks NW (upper left adjacent block), N (upper adjacent block), NE (upper right adjacent block), and W (left adjacent block) (lossless encoding on both the encoding side and the decoding side) This is the information you can know for).

That is, the prediction coefficients α _h and α _v are
α _h = (σ _h ^W + σ _h ^NW + σ _h ^N + σ _h ^NE + 4σ _h ') / 8 (13)
α _v = (σ _v ^W + σ _v ^NW + σ _v ^N + σ _v ^NE + 4σ _v ') / 8 (14)
As shown in the figure, set the value as close as possible as possible and perform linear prediction.

In equation (13), σ _h ^W is the horizontal correlation coefficient in the left adjacent block, σ _h ^NW is the horizontal correlation coefficient in the upper left adjacent block, and σ _h ^N is the horizontal correlation coefficient in the upper adjacent block. Also, σ _h ′ is a horizontal correlation coefficient default value such as 0.75.

In Equation (14), σ _v ^W is the vertical correlation coefficient in the left adjacent block, σ _v ^NW is the vertical correlation coefficient in the upper left adjacent block, and σ _v ^N is the vertical correlation coefficient in the upper adjacent block. Also, σ _v ′ is a vertical correlation coefficient default value such as 0.75.

[Fourth Invention Element ]
The residual transform signals according to the first to third invention elements are as follows when the one-dimensional linear prediction is repeated vertically and horizontally (according to algorithm 1), and when the two-dimensional linear prediction is performed (according to algorithm 2): The algorithm 4 is completely restored to a residual signal. Note that in the former case, the order of horizontal linear prediction and vertical linear prediction is reversed for encoding and decoding.

[Algorithm 3]
1. (* Vertical direction *)
2. for i ← 0 to 3
3. for j ← 0 to 3
4. p _{i, j} ← p _{i, j} + f (α _v p _{i, j-1} )
5. (* Horizontal direction *)
6. for j ← 0 to 3
7. for i ← 0 to 3
8. r _{i, j} ← p _{i, j} + f (α _h r _{i-1, j} )
[Algorithm 4]
1. for j ← 0 to 3
2. for i ← 0 to 3
3. r _{i, j} ← p _{i, j} + f (α _h r _{i-1, j} + α _v r _{i, j-1} -α _h α _v r _{i-1, j-1} )

  According to the present invention, it is possible to perform lossless video coding with high compression efficiency while improving efficiency by performing inter-frame prediction and maintaining affinity with the H.264 standard.

  Embodiments of the present invention will be described below with reference to the drawings.

FIG. 2 is a processing flowchart of lossless video encoding according to the third invention element . First, in the encoder, in accordance with the encoding method compliant with H.264, in step S101, each block in the image is sequentially predicted between frames or within a frame according to a prediction mode determined by the encoder.

In accordance with this prediction mode, horizontal and vertical prediction coefficients α _h ^mode and α _v ^mode are determined in step S102. Next, in step S103, the prediction residual signal R is obtained, and in step S104, the correlation coefficient (σ _h ^{W to} σ _v ^NE ) in the neighboring neighboring block of the block being encoded is expressed by the following equation (11), It calculates | requires according to Formula (12).

Next, in step S105, the horizontal prediction coefficient α _h and the vertical prediction coefficient α _v are determined, for example, as follows, following Expressions (13) and (14).

α _h = (σ _h ^W + σ _h ^NW + σ _h ^N + σ _h ^NE + 4α _h ^mode ) / 8 (15)
α _v = (σ _v ^W + σ _v ^NW + σ _v ^N + σ _v ^NE + 4α _v ^mode ) / 8 (16)
Using the obtained coefficient, linear prediction is performed on the prediction residual signal R by the algorithm 1 or algorithm 2 in step S106 to obtain a residual transformation signal P. Next, in step S107, this P is entropy encoded by a method compliant with H.264.

  In step S108, it is determined whether all the blocks have been encoded. As long as it is false, the process returns to step S101, and the encoding is repeated in the same manner for the next block.

FIG. 3 is a diagram showing an example of a functional block diagram of a lossless video encoding apparatus according to the third invention element .

The lossless video encoding device 1 includes an original image storage device 101 that stores an original image signal, an H.264 encoding device 102 that encodes the original image signal, and sequential blocks that sequentially generate position information of neighboring blocks adjacent to the block of interest. Position control unit 103, input control unit V104 that enumerates pixel pairs (a, b) vertically adjacent in the neighboring peripheral block of the block of interest, pixel pairs (a, b) adjacent to the left and right in the neighboring block adjacent to the block of interest Listed input control unit H105, product adder 106 that outputs Σab, adder 107 that outputs Σa and Σb, square sum adder 108 that outputs Σa ² and Σb ^2, and correlation coefficient calculation that outputs a correlation coefficient vessel 109, the correlation coefficient storage device 110 for storing the correlation coefficient, alpha _h ^mode and alpha _v alpha ^mode determining apparatus 111 determines the ^mode, alpha _h ^mode and alpha _v storing ^mode α _h ^mode, α _v ^mode SL Device 112, α _h, α _h for outputting α _v, α _v calculator 113, α _h, α _v stores α _h, α _v storage device 114, R memory for storing the prediction residual signal R for the target block 115, integers (i, j) ij loop controller 116 that generates a pair of, p _i, the two-dimensional linear predictor 117 to output a _j, P of i, p in j-th element _i, and stores the _j The P storage device 118 includes an encoded bitstream storage unit 119 for storing the encoded bitstream of P.

Α ^mode means a prediction coefficient for each prediction mode, α _h ^mode means a horizontal prediction coefficient for each prediction mode, and α _v ^mode means a vertical prediction coefficient for each prediction mode.

  The H.264 encoding device 102 reads an original image signal from the original image storage device 101 and performs encoding in units of blocks. Information on the “target block” position during encoding is sequentially transmitted to the block position control unit 103. The sequential block position control unit 103 sequentially generates position information of adjacent peripheral blocks (W, NW, N, NE shown in FIG. 1) of the block of interest.

  The input control unit V104 sequentially generates pixel value information in the 4 × 4 block at the position designated by the block position control unit 103 from the original image storage device 101 as a pair of the form (a, b). .

The order is as follows in equation (2):
(A, b) = (r _0,0 , r _0,1 ), (r _0,1 , r _0,2 ), (r _0,2 , r _0,3 ), (r _1,0 , r _{1 , 1), (r 1,1,} r 1,2), (r 1,2, r 1,3), (r 2,0, r 2,1), (r 2,1, r 2,2 ), (R _2,2 , r _2,3 ), (r _3,0 , r _3,1 ), (r _3,1 , r _3,2 ), (r _3,2 , r _3,3 )
And all adjacent pixel pairs in the top and bottom are listed.

Similarly, the input control unit H105 in equation (2)
(A, b) = (r _0,0 , r _1,0 ), (r _1,0 , r _2,0 ), (r _2,0 , r _3,0 ), (r _0,1 , r _{1 , 1), (r 1,1,} r 2,1), (r 2,1, r 3,1), (r 0,2, r 1,2), (r 1,2, r 2,2 ), (R _2,2 , r _3,2 ), (r _0,3 , r _1,3 ), (r _1,3 , r _2,3 ), (r _2,3 , r _3,3 )
And all the pixel pairs adjacent to the left and right are listed.

The product adder 106 receives (a, b) as input, obtains the product ab, and outputs Σab which is the sum of ab products for all 12 pairs. The adder 107 receives (a, b) as an input, and outputs Σa that is the sum of a and Σb that is the sum of b for all 12 pairs. Square sum adder 108 receives as input (a, b), obtains a product a ^2, b ^2, and? A ² is a square sum of all 12 pairs, and outputs the .SIGMA.b ² is b square sum .

  Correlation coefficient calculator 109 uses an adder / subtracter, a multiplier, a divider, and a square root calculator (not shown) from outputs from product adder 106, adder 107, and square sum adder 108.

を出力する。

Is output.

Correlation coefficient storage device 110, either by the control of the input control unit V104 or the input control unit H105, also, control of the sequential block position controller 103 W, NW, N, by either NE, sigma _h Remember ^W or σ _v ^NE .

On the other hand, the α ^mode determination device 111 appropriately determines α _h ^mode and α _v ^mode according to the encoding mode determined by the H.264 encoding device 102, and stores the α _h ^mode and α _v ^mode storage device. 112.

The α _h , α _v computing unit 113 performs addition (not shown) based on σ _h ^W to σ _v ^NE , α _h ^mode , α _v ^mode from the correlation coefficient storage device 110 and the α _h ^mode , α _v ^mode storage device 112. Α _h and α _v are output in accordance with Expressions (15) and (16) using a calculator and a divider, and stored in the α _h and α _v storage device 114.

  On the other hand, the H.264 encoding apparatus 102 generates a prediction residual signal R for the block of interest (4 × 4 size) during the encoding process. This is stored in the R storage device 115.

  The ij loop control unit 116 is a pair of integers (i, j) (0, 0), (0, 1), (0, 2), (0, 3), (1, 0), (1, 1). , (1,2), (1,3), (2,0), (2,1), (2,2), (2,3), (2,0), (3,1), ( 3,2) and (3,3) are generated in this order.

The two-dimensional linear predictor 117 obtains four pixel values (r _{i, j} , r _{i−1, j} , r _{i, j−1} , r _{i−1, j−1} ) from the R storage device 115, and also, alpha _h, alpha _v storage device 114 from the alpha _h, to give the alpha _v, a multiplier (not shown) these inputs, the adder, the rounding unit,
p _{i, j} = r _{i, j} −f (α _h r _{i−1, j} + α _v r _{i, j−1} + α _{h αv} r _{i−1, j−1} )
P _{i, j} is obtained according to the above and is output. The P storage device 118 receives this _{pi, j} and stores it in the i, jth element of P (specified by the ij loop control unit 116).

  The 4 × 4 residual transform signal P obtained in this way is entropy-coded by the H.264 encoding device 102 and stored in the encoded bitstream storage unit 119. The above processing is repeated until the input of the original image signal from the original image storage device 101 is completed.

FIG. 4 is a process flowchart of lossless video decoding according to the embodiment of the present invention . In step S201, the decoder sequentially entropy-decodes the residual transform signal P from the bit stream according to the H.264 method. Further, in step S202, horizontal and vertical prediction coefficients α _h ^mode and α _v ^mode are determined according to the encoding mode obtained from the bit stream.

In step S203, correlation coefficients (σ _h ^{W to} σ _v ^NE ) in the decoded neighboring neighboring blocks are obtained according to equations (11) and (12). Next, in step S204, the horizontal prediction coefficient α _h and the vertical prediction coefficient α _v are determined following Expressions (15) and (16).

  In step S205, the residual transform signal P is subjected to inverse linear prediction using the obtained coefficient, and the prediction residual signal R is obtained. Using this prediction residual signal R, in step S206, a decoded image is generated by a method compliant with H.264.

  In step S207, it is determined whether all the blocks have been decoded. As long as it is false, the process returns to step S201, and the decoding process is repeated until all the blocks are decoded.

FIG. 5 is a diagram showing an example of a functional block diagram of the lossless video decoding apparatus according to the embodiment of the present invention . The lossless video decoding device 2 includes an original image storage device 201 that stores an original image signal, an H.264 decoding device 202 that decodes an encoded bit stream, and sequential block position control that generates position information of neighboring blocks adjacent to the block of interest. Unit 203, input control unit V204 that enumerates pixel pairs (a, b) vertically adjacent in the neighboring peripheral block of the block of interest, and pixel pair (a, b) neighboring right and left in the neighboring peripheral block of the block of interest An input control unit H205, a product adder 206 that outputs Σab, an adder 207 that outputs Σa and Σb, a square sum adder 208 that outputs Σa ² and Σb ^2, and a correlation coefficient calculator 209 that outputs a correlation coefficient , the correlation coefficient storage device 210 for storing the correlation coefficient, alpha _h ^mode and alpha _v alpha ^mode determining unit 211 determines a ^mode, alpha _h ^mode and alpha _v storing ^mode α _h ^mode, α _v ^{mod e} storage device 212, α _h, α _h for outputting alpha _v, alpha _v calculator 213, α _h, α _h for storing alpha _v, alpha _v storage device 214, stores the residual transform signal P for the target block P storage device 215, an integer (i, j) ij loop controller 216 that generates a pair of, r _i, a two-dimensional inverse linear predictor 217 to output a _j, r _i, a _j of R i, j th element And an R storage device 218 for storing the encoded bit stream, and an encoded bit stream storage unit 219 for storing the encoded bit stream.

  The H.264 decoding apparatus 202 reads a bit stream from the encoded bit stream storage unit 219 and performs decoding in units of blocks. Information on the “target block” position during decoding is sequentially transmitted to the block position control unit 203.

  The sequential block position control unit 203 sequentially generates position information of neighboring peripheral blocks (W, NW, N, NE) of the target block. The input control unit V204 sequentially generates pixel value information in the 4 × 4 block at the position specified by the block position control unit 203 sequentially from the original image storage device 201 as a pair of the form (a, b). .

The order is as follows in equation (2):
(A, b) = (r _0,0 , r _0,1 ), (r _0,1 , r _0,2 ), (r _0,2 , r _0,3 ), (r _1,0 , r _{1 , 1), (r 1,1,} r 1,2), (r 1,2, r 1,3), (r 2,0, r 2,1), (r 2,1, r 2,2 ), (R _2,2 , r _2,3 ), (r _3,0 , r _3,1 ), (r _3,1 , r _3,2 ), (r _3,2 , r _3,3 )
And all adjacent pixel pairs in the top and bottom are listed.

Similarly, the input control unit H205 in equation (2)
(A, b) = (r _0,0 , r _1,0 ), (r _1,0 , r _2,0 ), (r _2,0 , r _3,0 ), (r _0,1 , r _{1 , 1), (r 1,1,} r 2,1), (r 2,1, r 3,1), (r 0,2, r 1,2), (r 1,2, r 2,2 ), (R _2,2 , r _3,2 ), (r _0,3 , r _1,3 ), (r _1,3 , r _2,3 ), (r _2,3 , r _3,3 )
And all the pixel pairs adjacent to the left and right are listed.

The product adder 206 receives (a, b) as input, obtains the product ab, and outputs Σab which is the sum of ab products for all 12 pairs. The adder 207 receives (a, b) as an input, and outputs Σa that is the sum of a and Σb that is the sum of b for all 12 pairs. Square sum adder 208 receives as input (a, b), obtains a product a ^2, b ^2, and? A ² is a square sum of all 12 pairs, and outputs the .SIGMA.b ² is b square sum .

  Correlation coefficient calculator 209 uses an adder / subtractor, a multiplier, a divider, and a square root calculator (not shown) from the outputs from product adder 206, adder 207, and square sum adder 208.

を出力する。

Is output.

Correlation coefficient storage device 210, either by control of the input control unit V204 or the input control unit H205, also, control of the sequential block position control unit 203 is W, NW, N, by either NE, sigma _h Remember ^W or σ _v ^NE .

The α ^mode determination device 211 appropriately determines α _h ^mode and α _v ^mode in accordance with the encoding mode (mode) determined by the H.264 decoding device 202, and stores it in the α _h ^mode and α _v ^mode storage device 212. To do. The α _h , α _v computing unit 213 performs addition (not shown) based on σ _h ^W to σ _v ^NE , α _h ^mode , α _v ^mode from the correlation coefficient storage device 210, α _h ^mode , α _v ^mode storage device 212. Α _h and α _v are output in accordance with Expressions (15) and (16) using a calculator and a divider, and stored in the α _h and α _v storage device 214.

  On the other hand, the H.264 decoding device 202 generates a residual transformation signal P for the block of interest (4 × 4 size) in the decoding process. This is stored in the P storage device 215.

  The ij loop control unit 216 includes integer (i, j) pairs (0, 0), (0, 1), (0, 2), (0, 3), (1, 0), (1, 1). , (1,2), (1,3), (2,0), (2,1), (2,2), (2,3), (2,0), (3,1), ( 3,2) and (3,3) are generated in this order.

The two-dimensional inverse linear predictor 217 obtains p _{i, j} from the P storage device 215 and outputs three pixel values (r _{i−1, j} , r _{i, j−1} , r _i−1) from the R storage device 218. _{, j-1} ), and α _h and α _v are obtained from the α _h and α _v storage device 214, and these inputs are obtained by a multiplier, an adder, and a rounding unit (not shown).
r _{i, j} = p _{i, j} −f (α _h r _{i−1, j} + α _v r _{i, j−1} −α _{h αv} r _{i−1, j−1} )
R _{i, j} is obtained according to the above and output. The R storage device 218 receives this r _{i, j} and stores it in the i, j th element (specified by the ij loop control unit 216) of the prediction residual signal R.

  The 4 × 4 prediction residual signal R obtained in this way is added to the prediction signal by the H.264 decoding device 202 and accumulated in the corresponding position of the original image storage device 201. The above processing is repeated until the bit stream from the encoded bit stream accumulation unit 219 is exhausted.

  The above lossless video encoding and lossless video decoding processes can also be realized by a computer and a software program, and the program can be recorded on a computer-readable recording medium and provided via a network. Is also possible.

In the above, specific examples of encoding according to the third aspect of the invention and decoding of the encoded data have been described as embodiments. However, other embodiments can be similarly implemented. It is clear from the above explanation.

It is a figure which shows the coded block adjacent to the attention block. It is a processing flowchart of lossless video coding. It is a figure which shows an example of the functional block diagram of a lossless video encoding apparatus. It is a processing flowchart of lossless video decoding. It is a figure which shows an example of the functional block diagram of a lossless video decoding apparatus.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Lossless video encoding apparatus 2 Lossless video decoding apparatus 101,201 Original image storage apparatus 102 H.264 encoding apparatus 103,203 Sequential block position control part 104,204 Input control part V
105, 205 Input control unit H
106, 206 Product adder 107, 207 Adder 108, 208 Square sum adder 109, 209 Correlation coefficient calculator 110, 210 Correlation coefficient storage device 111, 211 α ^mode determination device 112, 212 α _h ^mode , α _v ^mode storage devices 113, 213 α _h , α _v computing units 114, 214 α _h , α _v storage devices 115, 218 R storage devices 116, 216 ij loop control unit 117 two-dimensional linear predictors 118, 215 P storage devices 119, 219 Coded bitstream storage unit 202 H.264 decoding device 217 Two-dimensional inverse linear predictor

Claims (8)

A lossless video encoding device for lossless encoding of a moving image signal,
Means for calculating a prediction residual signal subjected to spatial prediction in intra-frame coding or temporal prediction in inter-frame coding for each block of an encoding target image signal;
Calculates the prediction coefficient used for linear prediction of the block being encoded from the correlation coefficient of the prediction residual signal in the encoded surrounding block and the prediction coefficient for each prediction mode determined according to the prediction mode of the block being encoded Means for storing in a storage device ;
Means for obtaining a residual transformation signal by performing spatial linear prediction on the prediction residual signal using a prediction coefficient stored in the storage device ;
Means for encoding the residual transform signal without quantizing the lossless conversion signal. A lossless video encoding method for losslessly encoding a moving image signal,
Calculating a prediction residual signal subjected to spatial prediction in intra-frame coding or temporal prediction in inter-frame coding for each block of an encoding target image signal;
A prediction coefficient of a block being encoded is calculated from a correlation coefficient of a prediction residual signal in an encoded surrounding block and a prediction coefficient for each prediction mode determined according to a prediction mode of the block being encoded, and a storage device Storing the steps in
The relative prediction residual signal, obtaining a residual transform signal by performing a spatial linear prediction using the prediction coefficients stored in the storage device,
Lossless video coding method characterized by a step of encoding without quantizing the residual transform signals. A lossless video decoding device for decoding an encoded stream encoded using the lossless video encoding device according to claim 1,
Means for inputting an encoded stream and decoding a residual transform signal for each block;
A prediction coefficient used for inverse linear prediction of the block being decoded is calculated from the correlation coefficient of the prediction residual signal in the decoded surrounding block and the prediction coefficient for each prediction mode determined according to the prediction mode of the block being decoded. Means for storing in a storage device ;
Means for obtaining a signal identical to the prediction residual signal at the time of encoding by performing spatial inverse linear prediction on the decoded residual transform signal using the prediction coefficient stored in the storage device ;
And a means for generating a decoded image from the prediction residual signal. A lossless video decoding method for decoding an encoded stream encoded using the lossless video encoding method according to claim 2 ,
Inputting an encoded stream and decoding a residual transform signal for each block;
A prediction coefficient used for inverse linear prediction of the block being decoded is calculated from the correlation coefficient of the prediction residual signal in the decoded surrounding block and the prediction coefficient for each prediction mode determined according to the prediction mode of the block being decoded. Storing in a storage device;
The decoded residual transform signals, and obtaining the same signal as the prediction residual signal in encoding by performing a spatial inverse linear prediction using the prediction coefficients stored in the storage device,
And a step of generating a decoded image from the prediction residual signal. A lossless video decoding method comprising: A lossless video encoding program for causing a computer to execute a lossless encoding process of an image signal of a moving image,
Said computer,
Means for calculating a prediction residual signal subjected to spatial prediction in intra-frame coding or temporal prediction in inter-frame coding for each block of an encoding target image signal;
Calculates the prediction coefficient used for linear prediction of the block being encoded from the correlation coefficient of the prediction residual signal in the encoded surrounding block and the prediction coefficient for each prediction mode determined according to the prediction mode of the block being encoded Means for storing in a storage device ;
Means for obtaining a residual transformation signal by performing spatial linear prediction on the prediction residual signal using a prediction coefficient stored in the storage device ;
Means for encoding the residual transform signal without quantization ;
A lossless video encoding program for functioning as A lossless decoding program for causing a computer to execute a lossless decoding process for decoding an encoded stream encoded using the lossless video encoding device according to claim 1,
Said computer,
Means for inputting an encoded stream and decoding a residual transform signal for each block;
A prediction coefficient used for inverse linear prediction of the block being decoded is calculated from the correlation coefficient of the prediction residual signal in the decoded surrounding block and the prediction coefficient for each prediction mode determined according to the prediction mode of the block being decoded. Means for storing in a storage device ;
Means for obtaining a signal identical to the prediction residual signal at the time of encoding by performing spatial inverse linear prediction on the decoded residual transform signal using the prediction coefficient stored in the storage device ;
Means for generating a decoded image from the prediction residual signal;
Reversible video decoding program to function as. A computer-readable recording medium recording a lossless video encoding program for causing a computer to perform lossless encoding processing of a moving image signal,
Said computer,
Means for calculating a prediction residual signal subjected to spatial prediction in intra-frame coding or temporal prediction in inter-frame coding for each block of an encoding target image signal;
Calculates the prediction coefficient used for linear prediction of the block being encoded from the correlation coefficient of the prediction residual signal in the encoded surrounding block and the prediction coefficient for each prediction mode determined according to the prediction mode of the block being encoded Means for storing in a storage device ;
Means for obtaining a residual transformation signal by performing spatial linear prediction on the prediction residual signal using a prediction coefficient stored in the storage device ;
Means for encoding the residual transform signal without quantization ;
A recording medium for a lossless video encoding program in which a program for functioning as a recording medium is recorded. A computer-readable recording medium recording a lossless decoding program for causing a computer to execute a lossless decoding process for decoding an encoded stream encoded using the lossless video encoding device according to claim 1,
Said computer,
Means for inputting an encoded stream and decoding a residual transform signal for each block;
A prediction coefficient used for inverse linear prediction of the block being decoded is calculated from the correlation coefficient of the prediction residual signal in the decoded surrounding block and the prediction coefficient for each prediction mode determined according to the prediction mode of the block being decoded. Means for storing in a storage device ;
Means for obtaining a signal identical to the prediction residual signal at the time of encoding by performing spatial inverse linear prediction on the decoded residual transform signal using the prediction coefficient stored in the storage device ;
Means for generating a decoded image from the prediction residual signal;
A recording medium for a reversible video decoding program in which a program for functioning as a recording medium is recorded.

Images