Paper Title
Feature-Based Robust 3-Drs Motion Estimation For High Definition Frame Rate Up-Conversion

Abstract
Frame Rate Up-conversion(FRUC) is essential to display a low frame rate to high frame rate video signals within the limited bandwidth. In this paper, we propose the FRUC of high definition(HD) video based on feature-based robust3-D recursive search(3-DRS) motion estimation(ME)by focusing the problem that 3-DRS algorithm can propagate some wrongly estimated motion vectors(MVs) between original frames. First, we set a feature point patch in previous frame through the features from accelerated segment test(FAST) detector by using adaptive thresholding process(ATP). Second, to calculate the initial MV, we implement the uni-directional ME from a feature point patch to corresponding block in current frame. Third, the weighted average value of feature point vectors is adopted for initial MV, then,perform the robust 3-DRS ME based on bi-directional by using initial MV to reduce the error propagation and get a true motion. Last, the overlapped block motion compensation(OBMC) is performed for getting better visual quality and reducing the blocking artifacts. Experimental results of proposed algorithm show that the performance subsequently improves that of conventional methods. Indexterms- Frame rate up-conversion, Motion estimation, Robust 3-D recursive search, Features from accelerated segment test.