METHOD OF COMPRESSING IMAGE SIGNALS BY VECTOR QUANTIZATION

A method of compressing image signals by vector quantization comprises the steps of dividing two-dimensional image signals into blocks B1 each comprising signals at M.times.N number of adjacent picture elements, selecting a vector that corresponds with the minimum distortion to the image signals in each block B1 from a code book comprising different vectors prepared by defining M.times.N number of vector elements, and encoding a number representing the selected vector to correspond to each block B1. The number representing the selected vector is defined to be larger or smaller as the distortion of the vector in the code book with respect to a predetermined vector is larger. Group signals of the numbers selected for the blocks B1 are divided into blocks B2 by putting together the group signals at the adjacent blocks B1, orthogonal transformation is carried out for the group signals in each block B2, and the transformed signals are encoded respectively with an intrinsic code length. Or, prediction encoding is carried out for the group signals of the numbers selected for the respective blocks B1.