Image enhancement for low bit rate wavelet-based compression

Abstract

At low bit rates, wavelet-based compression is advantageous over most block-based image compression schemes in that it does not generate visually objectionable blocking artifacts. However, the reconstructed images still suffer from loss of image details and sharpness, and coding artifacts, due to the loss of a considerable amount of information in the quantization. While in general it may be impossible to recover the information loss, some coding artifacts can be alleviated through the incorporation of image smoothness constraints using appropriate image prior model. Meanwhile, the constraints from the transform coding can be utilized to prevent further image degradation, such as oversmoothing of remaining image details. We propose a novel enhancement technique that utilizes a discontinuous image smoothness model to obtain good reconstruction in terms of both PSNR improvement and visual observation.