基于空-谱加权总变分的高光谱图像混合噪声去除算法

Hybrid Noise Removal Algorithm for Hyperspectral Images Based on Spectral Weighted Total Variation

通常的高光谱图像噪声模型是单一类型的,这不能尽量真实地反映高光谱图像噪声的真实情况.同时,一般地高光谱图像混合噪声的去除算法通常只考虑空间维,因此重构效果并不理想.对此,高光谱图像混合噪声存在一定对图像阅读和理解造成误判,因此需要进行降噪工作.通过引入空-谱加权总变分正则项,得到高光谱图像的空-谱加权总变分(SSWTV)模型,并利用分裂Bregman迭代算法对其进行快速求解.仿真和数值实验表明,该算法能够有效地在去除高光谱图像的单一噪声或混合噪声的同时,能够较好地保持恢复图像的边缘和细节信息.

The general noise model of hyperspectral images was a single type, which couldn＇t reflect the real situation of the noise of hyperspectral images. At the same time, the algorithm of removing the hybrid noise in the hyperspectral images generally considered the spatial dimension, so the reconstruction result was not ideal. In this way, a hyperspectral image reconstruction algorithm based on spatial- spectral weighted total variation was proposed. By introducing the space--spectral weighted total variation regulari- zation term,a spatial--spectral weighted total variation （SSWTV） model on hyperspectral image was ob- tained,and a split Bregman iterative algorithm was used to solve this problem quickly. Simulation and nu- merical experiments shown that the algorithm could effectively preserve the edge and detail information of the image while removing the single noise or mixed noise in hyperspectral images.