Video superresolution reconstruction based on subpixel registration and iterative back projection

摘要

To improve the spatial resolution of video, a superreso-nlution reconstruction method based on a sliding window is proposednutilizing the movement information between frames in the low-nresolution video. We propose a registration algorithm based on anfour-parameter transformation model through Taylor series expan-nsion, using an iterative solving method as well as the Gaussiannpyramid image model to estimate the movement parameters fromncoarseness to fine. Superresolution frames are reconstructed usingnan iterative back projection (IBP) algorithm. We also present thensuitable length of the sliding window and the reasonable iterationnnumber of the IBP algorithm in the video superresolution reconstruc-ntion. Our algorithm is compared to other algorithms on simulatednimages and actual color videos. Both show that our registration al-ngorithm achieves higher subpixel accuracy than other algorithms,neven in the case of large movements, and that the reconstructednvideo has better visual effects and stronger resolution ability. It cannbe extensively applied to the superresolution reconstruction of videonsequences in which the frames are different from each other mainlynby translation and rotation.