Spatial error concealment with sequence-aligned texture modeling and adaptive directional recovery

A spatial error concealment technique based on the sequence-aligned texture modeling and the adaptive directional recovery is proposed in this work. The sequence alignment technique captures the local variation and the global trend of image textures with surrounding uncorrupted pixels, and provides the best texture model under a given cost function. With the derived texture model, geometric interpolation is used to recover lost pixels adaptively based on pixel locations. There are four candidate modes of pixel sequences to recover lost pixels, and one of them is selected for the concealment purpose. The selection criterion is based on the texture pattern modes of surrounding uncorrupted blocks. The pixel sequences used for error concealment can be obtained from the computation of the decoder or the side information from the encoder. Extensive experimental results are given to demonstrate that our error concealment technique outperforms several benchmark methods in both objective and subjective tests.

Research highlights
► We propose a new spatial error concealment based on the sequence alignment algorithm and geometric interpolation.
► The sequence alignment algorithm estimates lost textures through the global and local trends of texture.
► Encoder can send the texture information of lost regions to decoder for better pixel recovery.
► With estimated texture information, our new error concealment performs pixel recovery with geometric interpolation.
► The proposed error concealment algorithm gives both competitive objective and subjective results comparing with the state of the art spatial error concealment algorithm.