A novel and fast feature based motion estimation algorithm through extraction of the background and moving objects

This paper presents a novel and fast Feature Based Motion Estimation algorithm which is developed for typical video-phone scenario. In essence it combines the technique of object extraction with traditional block based motion estimation methods by estimating the background and extracting the moving object continuously in the first stage, then performs a block based motion estimation on the extracted. Simulation of the algorithm with full search as core shows that the estimation time can be reduced by as much as 50%, while the MSE and PSNR remain almost the same as the full search results.