A Hermite Interpolation Based Motion Vector Recovery Algorithm for H.264/AVC

Abstract

Error Concealment is a very useful method to improve the video quality. It aims at using the maximum received video data to recover the lost information of the picture at the decoder. Lagrange interpolation algorithm is the most effective interpolation for error concealment while it lost some detail information. Hermite interpolation algorithm considers the change rate of the motion vector as well as the motion vector itself, which is more accurate. In this paper, we propose a novel method which uses hermite interpolation to predict the lost motion vectors. We take the change rate (derivative) of motion vector into account, and synthesizing the horizontal and vertical recovered motion vectors adaptively by the minimum distance method. The experimental result shows that our method achieves higher PSNR values than Lagrange interpolation.