Image signal compressing method by vector quantization

Claim 1

A method of compressing image signals by vector quantization, which comprises the steps of: (i) dividing two-dimensional image signals into blocks B wherein each of said block B is composed of the signals at M number of picture elements adjacent to one another, (ii) selecting a vector that corresponds with the minimum distortion to differences (x 1 -m, x 2 -m, x 3 -m, . . . , x M -m) between the image signals x in each of the blocks B and a representative value m of signals from a code book comprising a plurality of vectors different from one another and prepared in advance by defining M number of vector elements, and (iii) encoding the information representing the selected vector, and the information representing the representative value m, wherein the improvement comprises the steps of: (a) dividing said two-dimensional image signals into blocks B' whose picture element regions are deviated from the picture element regions of said blocks B, (b) calculating a representative value n with respect to the image signals in each of said blocks B', and (c) for each of partial regions R, whose picture element regions overlap the picture element regions of said blocks B', within each of said blocks B, using said representative value n, which is calculated for one of said blocks B' that includes one of said partial regions R, as said representative value m for said one of said partial regions R.

Claim 2

A method of compressing image signals by vector quantization, which comprises the steps of: (i) dividing two-dimensional image signals into blocks B wherein each of said block B is composed of the signals at M number of picture elements adjacent to one another, (ii) selecting a vector that corresponds with the minimum distortion to differences (x 1 -m, x 2 -m, x 3 -m, . . . , x M -m) between the image signals x in each of the blocks B and a representative value m of signals from a code book comprising a plurality of vectors different from one another and prepared in advance by defining M number of vector elements, and (iii) encoding the information representing the selected vector, and the information representing the representative value m, wherein the improvement comprises the steps of: (a) dividing said two-dimensional image signals into blocks B' whose picture element regions are deviated from the picture element regions of said blocks B, (b) calculating a representative value n with respect to the image signals in each of said blocks B', and (c) for each of partial regions R, whose picture element regions overlap the picture element regions of said blocks B', within each of said blocks B, using said representative value n, which is calculated for one of said blocks B' that includes one of said partial regions R, as said representative value m for said one of said partial regions R.

Claim 4

A method of compressing image signals by vector quantization, which comprises the steps of: (i) dividing two-dimensional image signals into blocks B wherein each of said blocks B is composed of the signals at M number of picture elements adjacent to one another, (ii) selecting a vector that corresponds with the minimum distortion to errors (x 1 -x 1 , x 2 -x 2 , x 3 -x 3 , . . . , x M -x M ) between the image signals x in each of the blocks B and interpolation-predicted values x for the image signals x from a code book comprising a plurality of vectors different from one another and prepared in advance by defining M number of vector elements, and (iii) encoding the information representing the selected vector, and the information utilized for calculating the interpolation-predicted values x, wherein the improvement comprises the steps of: (a) dividing said two-dimensional image signals into blocks B' whose picture element regions are deviated from the picture element regions of said blocks B, (b) calculating a representative value n with respect to the image signals in each of said blocks B', and (c) predicting said interpolation-predicted values x in each of said blocks B on the basis of said representative values n with respect to all of said blocks B' whose picture element regions partially overlap the picture element region of each of said blocks B.

Claim 5

A method of compressing image signals by vector quantization, which comprises the steps of: (i) dividing two-dimensional image signals into blocks B wherein each of said blocks B is composed of the signals at M number of picture elements adjacent to one another, (ii) selecting a vector that corresponds with the minimum distortion to errors (x 1 -x 1 , x 2 -x 2 , x 3 -x 3 , . . . , x M -x M ) between the image signals x in each of the blocks B and interpolation-predicted values x for the image signals x from a code book comprising a plurality of vectors different from one another and prepared in advance by defining M number of vector elements, and (iii) encoding the information representing the selected vector, and the information utilized for calculating the interpolation-predicted values x, wherein the improvement comprises the steps of: (a) dividing said two-dimensional image signals into blocks B' whose picture element regions are deviated from the picture element regions of said blocks B, (b) calculating a representative value n with respect to the image signals in each of said blocks B', and (c) predicting said interpolation-predicted values x in each of said blocks B on the basis of said representative values n with respect to all of said blocks B' whose picture element regions partially overlap the picture element region of each of said blocks B.