We presented a novel image filtering approach based on the shape-adaptive DCT transform (SA-DCT). Hard-thresholding and empirical Wiener filtering are performed in SA-DCT do- main, with an arbitrarily-shaped transforms support which is adaptively deÞned for every point in the image. The approach is used for the accurate denoising of grayscale as well as color images. Besides noise removal, the proposed method is also effective in dealing with those artifacts which are often encountered in block-DCT compressed images and videos.
Blocking artifacts are suppressed while salient image features are preserved. The luminance-driven shape-adaptive filtering can faithfully reconstruct the missing structural information of the chrominances, thus correcting color-bleeding artifacts.
The visual quality of the estimates is high, with sharp detail preservation, clean edges, and without unpleasant artifacts introduced by the fitted transform. The Pointwise SA-DCT algorithms demonstrate a remarkable performance, typically outperforming the best methods known to the authors.