Bayesian postprocessing algorithm for dwt-based compressed image
Wei Wen, Zhiyun Xiao, Silong Peng · 2005
The perceived quality of compressed images is severely degraded especially when the bit rate becomes very low. The traditional postprocess methods will lose their effect in dealing with the DWT-based compressed image at very low bit rate because they do not consider the blurring effect in quantization process. In this paper, we propose a new model for the postprocess by incorporating a blur kernel into it, which is used to deblur. Median filter is used to detect and penalize the quantization noise. Under Bayesian analysis, MAP estimation is given. Alternate iteration method is proposed to solve this problem. Numerical experiments show that the subjective perceived quality as well as objective evaluation is improved.