Evaluation of Compressed Images Based on Human Visual Characteristics and Its Application to Block Adaptive Quantization

Wataru KONDO, Norimichi TSUMURA, Yoichi MIYAKE
 
Department of Information and Image Sciences, Chiba University, 1-33, Yayoi-cho, Inage-ku 263-8522
 

Journal of The Society of Photographic Science and Technology of Japan, Vol.61, No.6 pp.343-351 (2001).

Abstract
Online-shopping became popular for vending various articles via computer network. The digital images of articles, however, do not have enough information such as actual 3D shape, texture, and color of objects. For recording objects as digital images without losing their color information, we consider that the spectral reflectance is the most desirable form. In this study, we discuss an evaluation of the image reproduced from those compressed multi-band images using image distortion map considering human visual characteristics. Furthermore, we discuss optimal compression method with block adaptive quantization based on the image distortion map. Large error areas on the image distortion map are finely quantized, while small error areas are roughly quantized. As a result, our proposed method reduces a compressed file size by about four percent compared with the conventional fixed quatization method.

Full paper ( to appeare )
[PDF] [PDF figures](in Japanese)

Back