高光谱图像条带噪声去除算法研究

Research on destriping algorithm for hyperspectral images

高光谱图像中往往存在严重的条带噪声,与常见噪声具有很大不同,现有的去噪方法对其不能完全适用。利用平滑滤波思想结合传统的矩匹配方法,对传统的矩匹配条带去除方法进行改进,对HJ-1-A卫星高光谱图像进行实验、研究,利用多种图像质量评价标准,将算法去条带后图像与传统的和已有的改进算法去条带后图像进行对比、评价,找出去条带效果好、原始图像信息保留能力强、适应性强的改进的矩匹配方法来应用于高光谱图像条带噪声的去除。

Hyperspectral images often have serious band noise.They are different from common noise.The existing denoising methods can not be fully applicable to them.The improved algorithm is proposed,which is combined with smooth filtering method and traditional moment matching method.Through experimental study on HJ-1-A satellite hyperspectral images and using a variety of evaluation criteria to compare,the best method to remove stripes of hyperspectral images is found.It can destripe effectively and retain the original image information well and have good applicability.