PURPOSE: To execute asymmetric coding not requiring extraction of 3-dimensional motion and a 3-dimensional structure.

CONSTITUTION: A converter 102 converts an image 151 received by an image input device 101 into a multiplex resolution space. A quantizer 105 applies vector quantization to a local pattern of an image 152 in the multiplex resolution expression outputted from the converter 102. A discrete depth calculation device 107 extracts information of a position and/or a rotary angle and/or a scale corresponding to each of plural code words included in an input image outputted from the quantizer 105 and information representing context of plural representative vectors. An algebraic coder 108 encodes the input image based on the extracted information.