The image processor of the present invention includes: an extracting means, an entropy decoding means, a combining means, and an inverse orthogonal transformation means. Coded data at respective stages in a certain block are extracted by the extracting means, and entropy-decoded by the entropy-decoding means, to obtain coefficient data at the respective stages in the block. The coefficient data at the respective stages are combined by the combining means to obtain coefficient data for the block. The coefficient data is subjected to inverse orthogonal transformation by the inverse orthogonal transformation means, to attain original image data for the block. The above series of processing are performed for all blocks, to attain an original image on the entire screen. The memory capacity required for combining the coefficient data at the respective stages may only be large enough to store coefficient data for one block.