(A) filter design for reducing blocking effect and ringing noise in block-based image coding

Abstract

Most of video coding standards generally use the block DCT and the quantization to compress the images. It is indispensable to originate the distortions of images called block effect and ringing noise, when the image is highly compressed. In this thesis, a filtering method, called line block filter(LBF) for reducing blocking effect and ringing noise is proposed without violating the video coding standard. In addition to the filter, error diffused quantization (EDQ) is used together for reducing the blocking effect. These LBF and EDQ can be used as a loop-filter and a post-filter. The LBF consists of two filters, such as SABF(Signal Adaptive Boundary Filter) and SAF(Signal Adaptive Filter). These filters are applied according to edge map which is made for detecting the contour. In the edge contour, no operation of the filters performs in order to preserve the detailed image. The SABF is for reducing the blocking effect in block boundary and SAF is for suppressing the ringing noise. Also the EDQ, which is used to improve the subjective quality of images with the filter in frequency domain, by diffusing the quantization errors to the near blocks. In simulation results, PSNR and bits rates produced by proposed methods are compared to those by the original coder. Each image made by all algorithm is illustrated together for subjective quality. Also the subjective quality is evaluated.