基于多尺度Retinex理论的遥感图像增强算法

多尺度Retinex算法是近年来常用的图像增强方法之一,针对该方法在处理遥感图像时存在过增强、灰度化和色彩失真等不足,结合I阶巴特沃斯函数良好的滤波特性,提出了一种改进多尺度Retinex理论的遥感图像增强方法.首先,采用I阶巴特沃斯函数取代多尺度Retinex中的高斯环绕函数,对降质的遥感图像R、G、B三个分量分别进行滤波操作;然后对滤波后的各通道下不同尺度的亮度图像分别进行加权和归一化处理,得到增强后的R、G、B三个分量;最后将三通道分量进行合成显示,得到最终的增强图像.文中采用中巴资源卫星(CBER-02)多光谱图像和日本ALOS卫星多光谱图像进行实验,结果表明,与传统的多尺度Retinex理论增强方法相比,改进后的方法在抑制图像灰度化、增强图像空间细节和降低图像颜色失真等方面有更好的实验效果.

Algorithm for Remote Sensing Image Enhancement Based on Multiscale Retinex Theory

Multi-scale Retinex（MSR）algorithm is a common method for image enhancement .However , the MSR has the disadvantages of over-enhancement ,gradation and color distortion in dealing with remote sensing images .Combined with the good filter characteristics of the I order Butterworth filter ,an improved MSR method for image enhancement was proposed .Firstly ,the Gaussian surround function in MSR was replaced by the I order Butterworth function .Three components R ,G and B from the degraded remote sensing image were filtered respectively .And then the luminance images at different scales for each channel were weighted and normalized ,obtaining enhanced R ,G and B components .Finally ,the enhanced image was obtained by synthesizing the three components .Multispectral images from the CBER-02 and ALOS satellites were tested ,Experimental results show that the improved MSR outperformed the traditional MSR in suppressing image gradation ,enhancing image space details and reducing color distortion .