Error concealment with directional filtering for block transform coding

In the paper, an error concealment algorithm using directional filtering for block-based coding is presented. Block transform coding is very powerful for image or video compression. Algorithms based on one of the most famous block transform coding methods, the discrete cosine transform, can provide good quality reconstructed video with higher than 20:1 compression ratio. As new applications arise, the problems of imperfect channels need to be considered. The problem addressed in the paper is the concealment of damaged regions in video images using the correlated information in the received signals. Spatial interpolation is an important approach to error concealment, especially in stressful image sequences where motion cannot be accurately predicted or in intra-coded frames which have no motion information. The authors propose a new approach which utilizes spatially correlated information thoroughly by performing interpolation based on a large local neighborhood of surrounding pixels. The use of directional information results in the recovery of edges, which have great influence in the perceived subjective quality of images. Computer simulation results have shown notable improvement over existing algorithms. If the damaged block is surrounded by validly decoded blocks, the edge recovery in the lost blocks is very successful