Error concealment for image transmission by multiscale Markov random field modeling

Transmission of visual data over communication channels introduces serious fidelity problems. The paper proposes an efficient error concealment algorithm for the image transmitted as encoded bitstreams. The proposed method introduces a multiscale Markov random field (MMRF) to model the original image. The established MMRF model is consequently employed to interpolate the missing blocks in the observing data via maximum a posterior (MAP) estimation. Compared with the conventional spatial interpolation scheme, this algorithm offers significant improvements on both subjective and objective qualities of the recovered image