Signal adaptive postprocessing for reducing blocking effects in JPEG image

Abstract

In this thesis, an algorithm is introduced to reduce the blocking artifacts of JPEG (Joint Photographic Experts Group) images. Reconstructed images from JPEG compression produce highly noticeable degradation near the block boundaries, called discontinuity adjacent blocks, because each block is transformed and quantized independently. Therefore, the reduction of blocking effects is a central issue in the post processing for better image quality. In order to reduce the blocking effects, we propose an algorithm consists of two stage enhancement procedures. At the first stage, an image is classified into two areas, i.e., an edge area and a relatively monotone area, according to the image intensity gradient by using a local adaptive threshold scheme. At the second stage, the staircase noise along the edge is smoothed by 1-D directional filter according to four directions of the edge, and the grid noise in the monotone area is smoothed by a signal adaptive lowpass filter. The comparison study between the proposed algorithm and other algorithms is performed computer simulation with several JPEG images. We represent some examples of processed images, PSNR (Peak Signal to Noise Ratio) figures, and histogram characteristics to demonstrate the significant improvement on the subjective and objective views.