In order to improve the spatial resolution of videos, utilizing the sub-pixel movement information between the low-resolution frames and the blur function of the imaging system, a blind video super-resolution reconstruction method is proposed. Firstly, through Taylor series expansion and least square solving method, the movement parameters between the adjacent frames in the sliding window are estimated from coarseness to fine. Secondly, according to the error-parameter curves generated through Wiener filter image restoration method, the parameters of the Point Spread Function (PSF) of the reference frame in the sliding window are estimated. Finally, super-resolution frames are reconstructed through Iterative Back Projection (IBP) algorithm. Experiments are performed on simulated low resolution images and standard test video and practical video, respectively. The results demonstrate the effectiveness of our approach and show the great importance of Gaussian blur estimation in video super-resolution reconstruction.
展开▼
