Compression Artifacts Removal Using an Adaptive POCS Algorithm and Explicit Region Modeling

An effective compression artifacts removal algorithm is proposed based on thin-plate spline interpolation and the theory of projections onto convex sets (POCS). The thin-plate spline model is used to model the luminance in the flat regions of the image. Adaptive projections are used to suppress the compression artifacts in the spatially active image regions. The algorithm includes block mapping, prediction of the spatial distribution of the quantization error and estimation of the visibility of the compression artifacts. Information obtained from the transform domain is used in the thin-plate interpolation phase. The adaptive projections use information from both the spatial and transform domains. Experiments performed with JPEG-compressed images demonstrate the effectiveness of the proposed algorithm in suppressing the compression artifacts, along with its ability for preserving the image sharpness.