Contrast Enhancement Framework for Suppressing JPEG Artifacts

Abstract

Contrast enhancement is a common tool in Computer Vision. It can be applied either globally, such as tone-curve adjustment, or locally, such as dehazing. The current algorithms show their limitations on compressed images because the compression artifacts are boosted at the same time. Applying JPEG artifacts reduction algorithms might not produce pleasant results. Pre-processing will over smooth the image. When applied as post-processing, the boosted artifacts cannot be reduced effectively. To resolve this problem, we introduce an automated method that integrates the artifacts reduction with the contrast enhancement algorithm. This method is based on one basic observation: when the image is decomposed into a structure layer and a texture layer, the JPEG artifacts are mainly contained in the texture layer. For the structure layer, we directly apply the contrast enhancement algorithm. Deblocking is applied to the texture layer. Finally combine these two layers in a reasonable manner.