A non-linear filtering for block-edge reduction in digital video

In digital video coding, the 2D block-based Discrete Cosine Transform (DCT) has been widely used, combined with the quantisation techniques of DCT coefficients, for the purpose of compression of video data. When coarser quantisation is employed, the DCT blocks become prominent in the reconstructed images and this affect is called block-edge artifact. Several linear filters have been proposed in the video coding literature to reduce this artifact, however, these techniques blur the block edges and thus reduces the signal-to-noise ratio of the filtered images compared to the reconstructed images. In this paper, a new non-linear filtering technique has been introduced and it reduces the block-edge artifact using the changes in local statistical characteristics of the pixels along the block edges. This technique reduces the block-edge artifact while minimising the blur along the block edges and thus gives an increased signal-to-noise ratio in the filtered images.