Phase-based side information generation in distributed video coding

In distributed video coding, the quality of side information (SI) directly affects the final compression efficiency and rate-distortion performance. In the process of generating SI, many methods require accurate pixel matching and produce intermediate frames based on calculating motion vectors. As a result, they perform poorly in scenarios such as light changes, motion blur, and so on. In this case, phase-based frame interpolation for video (PBFI) is proved to be effective. And it has the advantages of simple implementation and easy parallelization. However, this method cannot get a good result when interpolating the video sequence of large motion. This paper proposes a phase-based SI generation method, which applies PBFI to distributed video coding and improves it from two aspects. (1) Move the high frequency information to the low frequency, reduce the proportion of the high frequency components, so that more frequency components can be interpolated in phase; (2) By correcting the phase difference again, more accurate phase shift information can be obtained when facing the video sequence of large motion. Simulation results show that the proposed method can effectively improve the quality of the generated SI. Thereby the overall performance of the distributed video coding system can be improved.