An Improved Unsupervised Learning of Motion Estimation Based on Diamond Searching for Distributed Video Coding

Distributed video coding has received much attention in recent years. It shifts the complex motion estimation from the encoder to the decoder side, thus makes low-complexity encoding a reality. Among some motion estimation methods exploited for DVC, expectation maximization (EM) algorithm is the most effective. However, the full searching involved in EM algorithm makes it complex. So, in this paper, we adopt a diamond searching instead of the full searching to optimize the motion searching. The simulation results show that the proposed diamond searching saves time while with the almost same rate-distortion quality compared to the current EM algorithm.