For the predictive coding system using motion compensation, the vary of correction, i.e., the variety of vectors N, should preferably be as large as potential and the variety of image elements in a block be as small as possible. From this viewpoint, the method of in which essential hardware is decided by the value of N and a wide range of motion compensation disadvantageously needs an increased hardware. In distinction, the tactic of primarily makes use of M arithmetic models repeatedly for N occasions, and the hardware requirement is nearly decided by the value of M. Thus, it could be seen that the tactic of is extra advantageous. The tackle of the prediction sign generator memory 8-11a for writing the above-mentioned supplementary image components is ideally next to the addresses for storing picture parts that are essential for figuring out the optimum prediction perform for the current block, for facilitating read/write management. For this purpose, A1 by way of A8 are written following the addresses for storing x1 via x16 as proven in FIG. The essential storage capacity of the prediction signal generator memory 8-11a is expressed, using symbols k, L, m and n, as (2m+k)×(2n+L) for determining the optimum prediction perform of the present block, k×(2n+L) for supplementary picture parts for determining the optimum prediction operate of the next block, amounting to 2(m+k)×(2n+L) in whole.
In this case, image parts in a two-dimensional block across a quantity of scanning strains are rearranged, foxz168 ทางเข้า and film components in each block are transformed so that image components in one block continue to those within the subsequent block and all image elements in a two-dimensional block are aligned serially. Accordingly, presence of scanning lines needn't be of concern when finishing up prediction coding for picture parts in a two-dimensional block, and a two-dimensional block can be treated identically to a one-dimensional block produced from M picture elements on a single scanning line. As could be seen from the above, when scan conversion is carried out in advance, predictive coding by use of two-dimensional blocks can be not extra than predictive coding by use of one-dimensional blocks. According to the present invention, the above process is achieved for dedication of the optimum predictive function by repetitive use of M arithmetic models for N occasions, thus, hardware could be made small when the number of image components M included in a block is small. Moreover, the prediction error energy for a block for every operate can be obtained by single computation, permitting the selection of an prediction perform to be in contrast next whereas finishing up the computation. It implies that all of N-times computations are not always needed for figuring out the optimum prediction operate.
An adaptive coding system in accordance with declare 2 whereby said prediction operate for dedication of the analysis by stated second evaluation means is controlled in accordance with buffer reminiscence occupancy of the system. 11B reveals one instance whereby a differential vector sign produces twenty-five vectors foxz168 สมัคร on the x-y coordinates, by which digits in circles show the magnitudes, that's, the absolute values of the vectors. 11A, the pattern of the the size of variable word length code illustrated in connection with the first quadrant is utilized to the remaining quadrants but not depicted for simplicity of illustration. The above patterning of the length of variable word length code is grounded on the next. 4A, parts as designated by 2b, three, 4, 5 and 6 constitute a typical predictive coder which may schematically be illustrated as proven in FIG. 4B. The provision of the quantizer as proven at dotted line is not important.
On the opposite hand, after scan conversion has been carried out, the video signal is organized within the order of a11, a12, a21, a22, a13, a14, a23, a24 and so on, and computation for a two-dimensional block takes solely four picture element periods similar to for a11, a12, a21 or for a13, a14, a23 and a24, and generally M picture factor periods. When scan conversion is carried out upfront, computation for each two-dimensional block becomes the identical as that for a one-dimensional block including, for example, a11, a12, a13, and a14, thus significantly simplifying the computation control. Thus, prediction at a location shifted by -Vx within the x-direction and -Vy within the y-direction from a corresponding point in the earlier frame minimizes and optimizes the prediction error energy P. Therefore, the prediction operate expressed by use of the vector (-Vx, -Vy) is the optimum prediction perform, and is referred to herein as the movement vector. As described above, when a movement vector is decided in accordance with each motion vector data and prediction error data, the efficienty of coding could be remarkably improved particularly where the transmission velocity is low. Accordingly, the invention is particularly effective where a tv sign undergoes compressive coding and is transmitted at a low velocity.
11A. Further, when the quantizing attribute of the prediction error signal is made rough, image quality degradation inherent to the inter-frame coding referred to as "soiled window" occurs by which quantized noise stands still on the image so that as if an image were seen via a grimy foxz168 ทางเข้า window. On the other hand, within the movement compensation inter-frame coding, there arises a new drawback of degraded image quality by which the soiled window is seen shifting. The new drawback of degrading the image quality is outstanding at a standstill portion of the tv image picture particularly when the sign to noise ratio (S/N) of the enter tv signal is low. The purpose of occurrence of the degradation of the picture quality in the movement compensation inter-frame coding is as follows. Thus, as described above, in the movement compensation inter-frame coding, the predictive coding operation is effected by using a signal of the preceding frame having the highest diploma of similarity to be a signal in the blocks in the current body as a prediction sign. However, when the enter tv signal accommodates noise, even on the standstill portion, the degree of similarity to a shift vector showing that the blocks are standstill just isn't always the utmost with the end result that the standstill portion might inadvertently be judged as a transferring portion. Consequently, a grimy window which is standstill within the inter-frame coding is seen transferring in the movement compensation inter-frame coding, thus greatly degrading the picture quality.
Means for compressive coding of knowledge, stated info containing a minimal of data representing said optimum prediction perform and data representing a prediction error delivered out of said predictive coding means when said optimum prediction operate is used. The code compressor 7 has a knowledge fee matching operate which allows compressed information to be output at a continuing knowledge rate, and it features a prediction error signal encoder 7-1 linked between line 103a and a multiplexor 7-3, and a movement vector encoder 7-2 linked between line 108b and multiplexor 7-3 as shown in FIG. An adaptive coding system according to declare four whereby mentioned code length of stated variable word size code is minimum for mentioned inter-frame prediction and will increase because the norm of said prediction perform increases. An adaptive coding system according to declare 2 whereby said first analysis means evaluates stated prediction operate by encoding it into a variable word size code which has the minimal code length for the inter-frame prediction.
The principle of the movement compensation might be briefly described.