Retinex和小波变换去除遥感图像云雾方法分析

Analysis of the Methods to Remove Cloud from Remote Sensing Images Based on Retinex and Wavelet Theory

基于遥感图像中的云和景物信息频率分布特征,介绍了改进的Retinex方法和小波变换两种去云方法的基本原理,并结合实验分析两种方法的优缺点。Retinex算法对较暗区域的图像处理有明显效果,改进的Retinex方法则是基于经典Retinex,对具有较高亮度的遥感图像通过图像变换方法达到去云的目的。小波变换方法是将图片进行适当层次的小波变换,增大低层细节系数,突出景物信息,减小高层细节系数,适当减小近似系数,最后将所有系数重构,得到重构图像。实验结果评价及数据表明:基于小波变换方法优于Retinex方法。

This paper analyzes methods to remove cloud from remote sensing images based on the improved Retinex and wavelet transform. The method of Retinex is only effective on dark image, on which the improved Retinex algorithm is based by inversing image process, while the method of the wavclet theory is bascd on thc characteristics of the multi - resolution of the wavelet transform. The detailed coefficients in the low layers arc increased in order to stand out the scenery information, and the approximate coefficients propcrly are diminishcd, after the reconstruction of cloud is removed from the original images. The experi- ment results in that the algorithm of cloud reduction based on wavelet transform is better than Retinex.