基于亮度相关矩的MODIS和SPOT影像融合研究

MODIS and SPOT Images Fusion Based on the Intensity Correlation Moment

针对MOD IS影像空间分辨率较低的问题,提出了一种基于亮度相关矩的多分辨率图像融合方法。该方法首先对SPOT影像进行小波分解,将MOD IS影像构成的RGB颜色系统变换到IHS颜色系统;然后,根据强度分量和SPOT影像低频分量的均值和方差来定义图像亮度相关矩;最后,IHS逆变换和小波逆变换得到包含更多信息和有效特征的融合图像。试验结果证明该方法得到的融合图像在保留地物光谱信息和提高空间分辨率上都具有很好的效果。

In recent years, many methods for fusing panchromatic and simultaneously multispectral （SM） images have been developed, however, there is little work concerning the fusion of images with significantly different spectral properties and pixel spacings though existing techniques either cannot avoid distorting their spectral characteristics or introduce artifacts. As an application of the Intensity Hue Saturation（IHS） transform and the orthogonal wavelet decomposition（OWD） in order to enhance the spatial resolution of MODIS images, intensity correlation moment model（ICMM） has been proposed based on mean and the standard deviation of both the intensity component obtained by averaging three MODIS images and the approximate component of the SPOT image decomposed with an orthogonal wavelet basis to produce one new intensity component while minimizing the degree of resampling. This system uses a self-adaptive correlation response to determine that how much the corresponding wavelet coefficients contribute to the dominant features for accurately transferring visually chromatic emergent information from any number of the input images into a single fused image. Subsequently, the high spatial resolution MODIS images are achieved through reconstructing the fused OWD coefficients by using the modulated MODIS images obtained through an inverse IHS as the approximate component and the details of the SPOT image as high frequencies. By using the ICMM as fusion rule, texture and pattern details coming from the SPOT image can be modulated to MODIS images without altering their spectral properties and contrast because fusion in the correlation response domain significantly improves the reliability of the feature selection and information fusion process. This technique can be applied to improve spatial resolution for either colour composite or individual bands. By an image fusion experiment using MODIS and SPOT images of Dapeng bay area of Shenzhen, China, the visual evaluation and statistical analysis between the original MODIS images and the fused MODIS images confirm that ICMM overcomes the preservation tradeoff and can achieve better performance in terms of both preserving spectral information and improving spatial resolution of SM images. Finally, the influence of the decomposition levels on the result is also discussed.