Spatio-temporal super-resolution reconstruction based on robust optical flow and Zernike moment for dynamic image sequences

MeiYu Liang, Junping Du, Xiaoyong Li, Liang Xu, Honggang Liu, Yawen Li

Abstract: A novel and efficient spatio-temporal super-resolution reconstruction model based on robust optical flow and Zernike moment is proposed in this paper. The model does not rely on accurate estimation of subpixel motion, and is robust to noise and rotation. Moreover, it can effectively overcome the problems of block and hole artifacts in the traditional reconstruction methods. In our super-resolution process, first we propose an efficient robust optical flow motion estimation model based on the motion details protection, then introduce the bi-weighted fusion strategy to implement the spatio-temporal motion compensation, and finally apply the multi-frame information fusion scheme to make the spatio-temporal super-resolution reconstruction and optimization based on Zernike moment and non-local self-similarity. Experimental results demonstrate that the proposed method outperforms the existing methods in terms of both subjective visual and objective quantitative evaluations.