Adaptive Video Back Projection Super-resolution Method Using Non-local Prior

In this paper, a novel video back projection super-resolution method using non-local prior is proposed. Our approach has three steps. First, we make initial motion estimation with block-matching and fine block matching to search for the pixels with similar structural content from video sequences. In this process, an adaptive technique is introduced to avoid redundant search. Second, we do initial interpolation to the input video sequences and a bilateral filtering is applied to the interpolation frames to achieve edge-preserving image smoothing. Third, a video non-local means filter is applied to modify the error image in iteration step of the iterative back projection. The experimental result shows that the proposed method can increase image details, moreover, the chessboard effect and ringing effect along image edges can be removed and sharp and clear edges in visual perception will be obtained.