Real-time video postprocessing for deblocking and deringing on mediaprocessors

Blocking and ringing are two major artifacts in highly compressed images and videos coded by block-based discrete cosine transform. Many existing deblocking and deringing algorithms are computationally very expensive and/or cannot produce satisfactory results at very low bit rates. We have developed an adaptive deblocking algorithm and a clustering-based deringing algorithm. These algorithms can smooth out the blocking and ringing artifacts while preserving the strong edges and texture areas. In addition, our postprocessing algorithms have low computational cost. We have implemented them on a commercially available, programmable processor. For an image size of 352 X 288, our algorithms take only 3.17 ms, demonstrating the feasibility of real-time video postprocessing. (C) 2003 Wiley Periodicals, Inc.