ADAPTIVE FUZZY POST-FILTERING FOR HIGHLY COMPRESSED VIDEO (TP-L2)

Author(s) : Hao-Song Kong (Mitsubishi Electric Research Laboratories, USA) Yao Nie (University of Delaware, USA) Anthony Vetro (Mitsubishi Electric Research Labs, USA) Huifang Sun (Mitsubishi Electric Research Labs, USA) Kenneth Barner (University of Delaware, USA)

Abstract : This paper presents a newly developed adaptive post-filter for removing blocking and ringing artifacts in the highly compressed image and video. The proposed post-filter adaptively adjusts its filtering window size and the spread parameter in the fuzzy membership function according to the variance value in an edge map. Under the guidance of the edge map, the post-filtering operation is directly applied to the edge blocks while keeping smooth and textured blocks unaltered. The experimental results demonstrate that the proposed method possesses a good edge preserving property and can maximally remove coding artifacts. Compared with the most popular referenced methods, the proposed post-filter achieves better subjective quality.

Menu