Spatial Error Concealment Using Directional Interpolation and Weighted Averaging(DIWA)

Abstract

In this paper, we propose a new spatial error concealment method which uses the direction of the edges when a block is lost due to error in an I-frame of H.264 image. The method extracts the edges from the partitioned groups of adjacent blocks. When the edges are not clear and the edge clearness is below the pre-defined threshold, it ignores edges. If edges are extracted, it finds the spatial direction vectors. Each pixel in the lost block is restored by using weighted interpolation in which the weight is dependent on the location and edge direction vectors. When no edges are extracted, weighted average is used to restore each pixel. The proposed method can restore piwels for a macroblock or a slice unit. Our method shows image quality improvement of 0.92 dB to 0.11dB when compared to several dther existing methods.