Frame rate up-conversion based on matching classification and dominant object tracking

Abstract

Frame rate up-conversion (FRUC) is effective to enhance video quality at LCD TV. In this paper, a new method based on matching classification and dominant objects tracking is proposed. Blocks to be interpolated are classified into 4 types: matched, unmatched, uncertain and unrelated according to the results of bilateral block matching. Dominant object tracking helps to refine the results. Then an adaptive median motion vector (MV) filter based on the classification results is employed to smooth the MV field. Empirical results show that the proposed method provides high quality MV field and improves the performance of FRUC.