Optical Flow Based Dc Surface Compensation for Artifacts Reduction

Xin Jin,Satoshi Goto,King Ngi Ngan

DOI: https://doi.org/10.1109/pcs.2009.5167414

2009-01-01

Abstract:The block-based prediction, transform and quantization reduce the redundancy in video data efficiently. However, the blocking artifacts are introduced because of quantization error and prediction scheme. Even if the de-blocking filter is applied like that designed in H.264 and AVS, the blocking artifacts are still visible especially in videos with smooth area coded at low bit-rate. In this paper, an optical flow based technique is proposed to further reduce the blocking artifacts by compensating the DC surfaces of the decoded images using joint optimization of optical flow and edge difference. The proposed algorithm automatically classifies the macroblocks in the decoded images according to their texture feature and endures the smooth area with smooth brightness variation by minimizing the joint function of optical flow magnitude and edge smoothness. It produces a more pleasing image without over-filtering, which can be further applied as an in loop filter for video coding.