A Robust Video Super-resolution Based on Adaptive Overlapped Block Motion Compensation

This paper has proposed an improved video super-resolution algorithm to reconstruct high quality pictures in a low resolution video sequence from existing high resolution key frames. Firstly, UMHexagonS algorithm search pattern is performed to obtain motion vectors between the up-sampled low resolution frame and key high resolution frames. Then, motion vectors are classified and refined by motion vector classification process. Finally, low resolution frames are super-resolved by adjacent key high resolution frames. Experimental results illustrate that the proposed method has good performance both on objective PSNR value and subjective visual effect especially with big and irregular motions videos.