An Improved Hyperspectral Image Super Resolution Restoration Algorithm Based on POCS.
Yulei Wang,Xinxin He,Yao Shi,Qingyu Zhu,Haoyang Yu
DOI: https://doi.org/10.1109/igarss47720.2021.9553830
2021-01-01
Abstract:Super-resolution reconstruction is a rapidly growing research area in hyperspectral data processing. However, there exist some problems, such as edge blur, burr in the smooth area, subjective design of iteration times, et al. This paper analyzes the causes of blur and burr, and puts forward some countermeasures according to these problems. Firstly, gradient interpolation is used instead of the traditional nearest neighbor interpolation, which alleviates the edge blur to a certain extent problem, the projection operator calculated from the gradient map is introduced into the projection formula to solve the burr phenomenon in the smooth area. Then, the mean square error of the reconstructed image of adjacent iterations is used to measure the similarity of the reconstructed image between two adjacent iterations, which is used as the stopping criterion of iterations, avoiding the subjectivity of setting the iteration times artificially. Finally, the proposed algorithm is applied to every band of hyperspectral image. Experimental results show that the proposed algorithm has better performance than the traditional POCS algorithm in visual effect and quantitative criteria.