Spectral Information Recovery for Compressed Image Restoration

In this paper, we attempt to solve the compressed image restoration by considering the recovery of spectral information in compressed images. For each NxN block in images, we convert the recovery problem to a route searching process in the N²-dimensional vector space R^{N 2} spanned by all DCT coefficients. The quantized DCT coefficients vectors A capped is known as the starting node of the route, while the target is original DCT coefficient vector A. In order to access A from A capped, a recovery route has to be built. Total variation (TV) based regularization can produce a group of estimated nodes by dynamic adjustment on images in spatial domain.