ベクトル量子化を用いたサブバンド画像符号化

Abstract

Image subband decomposition enables efficient coding matched to the statistics of each subband and human visual characteristics. Vector quantization (VQ) provides a powerful means of bit rate reduction taking advantage of remaining intra and inter band correlation of the decomposed subband images. This paper describes image coding schemes which combines subband decomposition and VQ for still pictures and video sequences. For still picture coding, we propose "same orientation" inter-subband VQ with vector power "thresholding" and "subband limitation". For video sequence coding, we propose subband VQ of motion compensated (MC) prediction difference. Computer simulation results present that 1) "thresholding" and "subband limitation" are very effective for low bit rate coding of still pictures and 2) "same orientation" inter-subband VQ of MC prediction difference shows higher performance than intra-subband VQ for video sequences.