A Study Of SAR And TM Image Fusion Based On Local Statistical Features Of Wavelet Transform Coefficients

Synthetic Aperture Radar (SAR) image and TM image are the major data sources in remote sensing application domain. Because of the difference of their imaging mechanism, there are complementary informations between the two images. The image fusion technique can integrate them into one image learning from their strong points to offset their weakness, and comparing with the single data, the fusing result is more excellent. It has a significant meaning for the builduping of the image definition and interpretation, and improving the precison of object identification and the reliability of classification.After analyzing the principles of image wavelet decomposition and composition, the local statistical features of WT coefficients are summarized, which shows that there are relativities not only between the WT coefficients of one image, but between the WT coefficients of different images that are of the same objects. In the course of SAR and TM image fusing, the fusion rule is determi-need by the relativities of WT coefficients, and the fusion factors of high frequency are based on the local statistical features such as local deviation and local correlativity, thus the fusion results are better than the traditional one which is based on the single pixels. The strongpoint and validity of this method can also be verified through the subjective and objective analyzing of fusion images.The fusion methods based on local deviation or local correlativity have advantages each in keeping the detail information and spectrum character. A improving algorithm is bring forward, which combines the strongpoint of two methods above. And the quality of this fusion image is more ideal especially at definition. It has an applid foreground.