A highly effective DCT-domain filtering algorithm for deblocking transform compressed images
Song Zhang, Fang Zhu · 2008
In this paper, a new DCT-domain filtering technique is proposed to remove the blocking artifacts in highly compressed images. We choose overlapping cross-block-boundary 16times16 macro-block as basic processing unit to ensure effective deblocking. For each macro-block, simple adaptive low pass filtering (LPF) is performed, which generalizes an earlier scheme introduced in. Experimental results show that the new method does a good job and outperforms several classic DCT-domain deblocking algorithms both in terms of objective and subjective measures. Computational issues with the new scheme are also briefly discussed. Moreover, the proposed system is shown to provide flexible framework to include more adaptive operations.