Image coding apparatus

Abstract

An image coding apparatus wherein an image signal including a locally high signal can be coded efficiently with minimized distortion by adaptively performing quantization for each subblock. An input image signal is divided into a plurality of subband signals by a subband division section. The subband signals from the subband division section are each divided into a plurality of subblocks by a subblock division section. In a classification section, a subband signal is classified, for each of the subblocks from the subblock division section, in terms of a feature parameter of an image, for example, a dynamic range, and a class index is outputted. In a quantization section, a quantizer is changed over for each subblock in response to the class index from the classification section. Each subblock is coded with a necessary minimum number of bits. The bit number is multiplexed with a code. Accordingly, the code can be decoded by referring to the bit number.