Multiple Features Matching Based Bidirectional Motion Estimation for Frame Rate Up-conversion
Xue Chunlin · Video Engineering · 2015
It usually results in the quality degradation of the interpolated frame in frame rate up-conversion that lots of periodical repetitive patterns in the texture region of video frame lead to the mismatch of video blocks for bidirectional motion estimation. To overcome this problem,this paper proposes to extract multiple features of video frame and add them into the process of block-matching,and thus the probability of appearing mismatch is lowered. Since the video sequence naturally contains color information,the chrominance component is firstly mixed into the block-matching. Besides,human's eyes is obviously sensitive to image edges,and therefore the gradient component,which is computed by the simple Sobel operator,is used to reveal the edge feature and mixed into the block-matching. Although the multiple features matching can effectively improve the accuracy of motion estimation,it introduces also the higher computational complexity. To reduce the computational complexity of multiple features matching,a special template is designed to combine multiple features into a single panel,thereby completing multiple features matching by performing only a block-matching operation,which improve the accuracy of motion estimation while guaranteeing a low computational complexity. Experimental results show that the proposed algorithm can improve the both subjective and objective quality of the interpolated frame with a low computational complexity.