Image recompression and perceptual distortion analysis
Keesook J. Han, Mark A. Robertson, Bruce W. Suter · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2004
The aim of this research is to recompress the JPEG standard images in order to minimize the storage and/or communications bandwidth requirements. In our approach, we convert existing JPEG images into JPEG 2000 images. The proposed image restoration method is applied to improve the visual quality when the bit rate becomes low and visually annoying artifacts appear in existing JPEG image. The JPEG restoration algorithm here makes use of the DCT quantization noise model along with a Markov random field (MRF) prior model for the original image in order to formulate the restoration algorithm in a Bayesian framework. The maximum of a posteriori (MAP) principle based convex model is applied to restore images. The restored image is then compressed with the JPEG2000. The cumulative distribution function (CDF) based visual quality metric method has been developed to measure coding artifacts in large JPEG images. Perceptual distortion analysis is also included in this paper.