Deblocking algorithm in MPEG-4 video coding using block boundary characteristics and adaptive filtering

Abstract

In this paper, we propose an efficient deblocking algorithm that uses block boundary characteristics and adaptive filtering in spatial domain. In the proposed algorithm, we detect the block boundary with blocking artifacts and classify the detected block boundaries into smooth regions or complex regions based on the statistical characteristics of neighborhood block. Thereafter, spatial adaptive filtering is performed. In smooth regions, the blocking artifacts seem a step-wise function, so the blocking artifacts can be reduced by a simple nonlinear 1-D 8-tap filter. In complex regions, there exist the blocking and ringing artifacts, so we propose adaptive filtering based on a feedforward neural network to reduce the blocking and ringing artifacts simultaneously. The horizontal and vertical block boundaries are processed separately with a different neural network. Experimental results show that the proposed algorithm gives better results than those of the conventional algorithms from both a subjective and objective viewpoints.