Research and VLSI Implementation on De-blocking Method for MPEG

Data compression has developed for decades, it is used in many different areas, one of which is image processing. MPEG as we all know, is the name of a working group focusing on image processing technology, and it is also one of the popular standard among all image compression. The video encoding/decoding of MPEG has great compression percentage which can keep a good image quality at the same time. So it is widely used in image processing field. But meanwhile, blocking artifact noise comes along with compression as an unsatisfying “extra-effect”. In order to suppress the blocking artifact and improving the quality of images or video, many algorithms of de-blocking are proposed these years. However, most of existing algorithms are not suitable for the implementation on video-processing chips, because their great amount computation would eventually lead to high cost in chips. To balance the computation and cost when integrating into single chip, the adaptive post-processing algorithm is proposed by H.-S. Kong et al. This algorithm can effectively reduce the blocking artifact as well as reduce the computation. But it is insufficient to preserve the edge information of images. Based on the algorithm that H.-S. Kong et al proposed, a post-processing de-blocking algorithm for MPEG video stream is brought to all in this paper. This method processes filtering based on the classification, which takes the circumscription of the two 4 by 4 areas near the block borderlines(vertical or horizontal). It determines the result out of three classification by calculating the pixels value of the two 4 by 4 areas during the procedure, which are smooth gradual changing area, luminance step area and complex texture area. Then it decides proper solution to use among strong filter, weak filter and non-filter according to the classification. This method offers a better solution to the problem of de-blocking while preserving edge information, and yet has a smaller calculating amount to meet the low complexity image de-noising requirement of video-processing chips. It has a good applications perspective on VLSI implementation.