Content-Adaptive Interpolation for Spatial Error Concealment

When transmitting encoded images over a communication channel, the reconstructed image quality can be substantially degraded by channel errors. This paper presents a spatial error concealment algorithm that utilizes variance of surrounding pixels, and then classifies each error block (EB) into two categories: uniform block and edge block. For uniform block, nearest border prior spatial interpolation is adopted to restore missing pixels. We use the Wiener interpolation algorithm and special interpolation sequence for edge block. Experimental results indicate the proposed algorithm can attain well restored quality of intra-frames both subjectively and objectively. Meanwhile, the computational cost of the proposed algorithm can be significantly reduced, compared to Lipsilas method. And the restored quality is almost the same, sometimes even much better.