An adaptive POCS algorithm for compression artifacts removal

Abstract

An effective compression artifacts removal algorithm is proposed based on the theory of projections onto convex sets (POCS). It includes a block classification procedure, a ringing detection procedure, prediction of the spatial distribution of the quantization errors and estimation of the visibility of the compression artifacts. Information gained from both the spatial and transform domains, is incorporated into adaptive projections. Experiments performed on JPEG-compressed images, demonstrate the effectiveness of the proposed algorithm in suppressing both blocking and ringing artifacts, as well as the ability of the algorithm to preserve the image sharpness.

Effective Decompression of JPEG Document Images

“…Original signal model Domain MAP Gibbs [7], [17], [18], BS-PM [19] DCT & Spatial POCS Neighboring constraint sets [6], [8], [9] DCT & Spatial TV Gradient magnitude sum [10], [20], [21] DCT & Spatial SRLD…”

“…Traditionally, POCS-based theory [6], [8], [9] has been employed for post-processing image coding artifacts. In its essence, a POCS-based method defines a set of constraints, each of which is described by a closed convex set.…”

“…Depending on applications, other constraints can be established to describe the smoothness of the original image. Such constraints, as defined in [8], [9], for instance, account for the close correlation between two adjacent pixels in either the horizontal or vertical direction. Because a sufficiently large number of constraints must be created to characterize the original image, one of the defects of the POCS-based method involves a variety of parameters used to define these constraints.…”

“…Iterative methods: Typical iterative methods for artifact post-processing include maximum a posteriori estimation (MAP) [7], [17]- [19], projection onto convex sets (POCS) [6], [8], [9], total variation (TV) regularization [10], [20], [21], and sparse representation based on a learned dictionary (SRLD) [11], [13], [14]. Table I shows the main characterizations of each method, and the details are presented hereafter.…”

Scale-space method of image ringing estimation

“…The size of this neighborhood is fixed a priori. In [7], the presence of ringing effect is detected by comparing the directions of image gradients at different scales.…”