Shape-adaptive DCT for denoising and image reconstruction

摘要: The shape-adaptive DCT (SA-DCT) can be computed on a support of arbitrary shape, but retains computational complexity comparable to that the usual separable block DCT. Despite near-optimal decorrelation and energy compaction properties, application SA-DCT has been rather limited, targeted nearly exclusively video compression. It recently proposed by authors8 employ for still image denoising. We use in conjunction with directional LPA-ICI technique, which defines shape transform's pointwise adaptive manner. thresholded or modified coefficients are used reconstruct local estimate signal within adaptive-shape support. Since supports corresponding different points general overlapping, estimates averaged together using weights depend region's statistics. In this paper we further develop novel approach extend it more restoration problems, particular emphasis deconvolution. Simulation experiments show state-of-the-art quality final estimate, both terms objective criteria visual appearance. Thanks support, reconstructed edges clean, no unpleasant ringing artifacts introduced by the fitted transform.