A New RS Image Fusion Method Based on Feature

Abstract

The image feature fusion is a kind of data fusion that is based on spectral structure and texture feature of the objects. This paper proposed feature fusion after enhancing image edge with wavelet's multi-resolution analysis to largely improve image's definition and resolving power. In the fusion process, instead of using the conventional HIS-Wavelet fusion, we directly introduced RGB-Wavelet transformation which decomposed RGB bands of the multi-spectral image and high-resolution image by wavelet separately, then used low frequent parts of the R,G,B bands and high frequent parts of the high-resolution image to do the image fusion, finally had new R,G,B bands as the new fusion image. In the way, it not only improved definition and resolution of the multi-spectral image, but also retained the color feature of multi-spectral image which is important for multi-spectral fusion. At last we validated the conclusion with an experiment. This method was based on wavelet multi-resolution analysis and RGB color space, so it didn't generate any computing error of color space transform. Compared the fusion results of RGB-wavelet and HIS-wavelet methods, it could find that not only precision of visual estimation but also guide lines of quantitative analysis such as definition, space resolution had greater improvement.