Video error concealment using Speeded Up Robust Features and affine transformation

Error concealment techniques try to mitigate the effect of channel error, using spatially and/or temporally neighboring macroblocks. This paper proposes a novel method for temporal error concealment. The change from one frame to the next frame is modelled as an affine transformation and Speeded Up Robust Features (SURF) algorithm is used to find correspondence between present and past frame. The set of matching points between the two frames is used to find an affine transformation. Random Sample and Consensus (RANSAC) algorithm is used to effectively remove outliers. The past frame is then affine transformed using the transformation matrix obtained. Corrupted regions are concealed using Temporal Replacement (TR) in the current frame with the transformed past frame as reference. Experimental results established the efficacy of the proposed algorithm over other temporal error concealment scheme in terms of both objective and subjective measurements.