基于MATLAB的IHS变换与小波变换影像融合与应用

Method of Fused Image Based on IHS and Wavelet Transform and Realization in MATLAB

为了提高Landsat系列卫星多光谱影像的目视解译效果,提出了一种基于IHS变换和小波变换相结合的影像融合方法。在Matlab开发环境下,将多光谱影像的亮度分量I与全色影像先进行直方图匹配,以加大两者的相关性,减小融合影像的光谱偏差,然后分别进行小波分解,最后按照一定融合规则进行影像融合。融合规则中最重要的是小波基、小波分解层数和小波系数的选择。通过比较不同的小波基函数,在不同分解层数下的统计参数值及目视融合效果,最终确定选择coif5作为小波基函数,进行三层小波分解。小波高频系数的选择采用区域的标准差法,选取以当前处理像元为中心的局部区域标准差最大的影像小波系数作为融合影像对应的小波系数。从融合后影像看,本文提出的方法要优于单一的IHS或小波变换融合方法,道路、河流、农田及林地等纹理及细节信息都较融合前明显提高,同时,较高的相关系数和较低的光谱扭曲度,表明光谱信息也保留很好。实验表明,将ETM+多光谱影像和全色影像融合的方法是可行的,融合后影像不仅空间分辨率明显提高,而且又较好地保留了多光谱影像的光谱信息特征。

In order to improve eye-interpretation effect of multi-spectral images of Landsat series, an integrated technique based on IHS and wavelet transformation is proposed in this paper. For increasing relativity and decreasing spectral deviation, firstly, under the developing environment of Matlab, the intensity component of multi-spectral image and the panchromatic image histogram wavelet, finally fuse images by special rules. bases, ties out wavelet decomposed layers and wavelet repeated match each other, and then decompose respectively using Among the rules, the most important thing is the choice of wavelet coefficients. In order to choose the best parameters, the author carexperiments using assessment parameters ： mean value, standard deviation, average gradient, correlation coefficient and spectral distortion. Compared with other parameters, wavelet base of coifs made better fusing result in this paper. Therefore, parameters are fixed on coif5 and decomposed of three layers, and reconstructed coefficients based on regional standard difference. The experiment indicates that the proposed method is much better than single transform of IHS or wavelet. Observed from the fused image, the lines and details of road, river, farm and forest are improved obviously. At the same time, the higher correlation coefficient and the lower distortion show that spectral spectral information is kept down well. The experiment through fused ETM ＋ multi-spectral and panchromatic images verified the method of the author proposed is feasible, the result is not only improved obviously in spatial resolution, but also preserves multi-spectral information effectively.