Image super-resolution reconstruction based on NEDI constraints
Xinxin Chen, Huiquan Wang · 2024
The classical NEDI algorithm causes covariance parity distortion among multiple original images due to the different structures of reference pixel points in the covariance estimation window, resulting in edge artifacts and blurring phenomena in the reconstructed images. The solutions of optimal window selection and transforming the edge consistency constraints into a priori conditions are proposed to address this problem and are experimentally verified in single-frame image reconstruction and sequence image reconstruction. The experimental results show that the algorithm in this paper solves the problems of jumping points and blurring at the edges of the reconstructed image caused by the classical NEDI algorithm's pairwise distortion, and has a better visual effect.