Wavelet Pyramid Image Coding with Predictable and Controllable Subjective Picture Quality (Special Section on Information Theory and Its Applications)

概要

論文の詳細を見る
A new method by which images are coded with predictable and controllable subjective picture quality in the minimum cost of bit rate is developed. By using wavelet transform, the original image is decomposed into a set of subimages with different frequency channels and resolutions. By utilizing human contrast sensitivity, each decomposed subimage is treated according to its contribution to the total visual quality and to the bit rate. A relationship between physical errors (mainly quantization errors) incurred in the orthonormal wavelet image coding system and the subjective picture quality quantified as the mean opinion score (MOS) is established. Instead of using the traditional optimum bit allocation scheme which minimizes a distortion cost function under the constraint of a given bit rate, we develop an "optimum visually weighted noise power allocation" (OVNA) scheme which emphasizes the satisfying of a desired subjective picture quality in the minimum cost of bit rate. The proposed method enables us to predict and control the picture quality before the reconstruction and to compress images with desired subjective picture quality in the minimum bit rate.
社団法人電子情報通信学会の論文
1993-09-25