A method and system are disclosed for compressing motion image information by
means
of predictive coding with a high compression ratio while maintaining high image
quality. An image in each frame is divided into blocks in advance, and each of
all blocks is approximated by (replaced with) a single plane defined by three data
representing the magnitude of the intensity of the block, the block-to-block gradient
of the intensity in the x direction, and the block-to-block gradient of the intensity
in the y direction. Furthermore, corresponding pixels are compared with each other
in each block consisting of nm pixels between frames, and differential information
is generated according to the comparison result. Pixels indicated as being not
greater than a parameter is treated (deleted) as pixels having no difference between
frames. Information compression is performed for pixels indicated as being greater
than the parameter.
