A new image fusion strategy based on target segmentaion

Abstract

By analyzing the characteristics of target in multi spectral image and panchromatic image, a new fusion strategy based on target segmentation is proposed to fuse multi spectral image and panchromatic image. Firstly, Hue-Saturation-Intensity (HSI) transform is performed on the multi spectral image. Secondly, the intensity component by HSI transform is segmented by the expectation maximization (EM) algorithm and the panchromatic image is segmented by fuzzy C means (FCM) clustering algorithm, and obtains the better target area, followed by effective filling of the target area to get the new intensity component. Finally, the new fusion image is obtained by inverse HSI transform. Compared with the traditional HSI method, experiment results shows that the proposed strategy not only increases information entropy and average gradient of fusion image, but also decreases spectral bias index. Therefore, the proposed strategy is better than traditional HSI method. It not only enhances spatial resolution of fusion image and obtains detailed and feature information, but also preserves spectral information of the original multi-spectral image well.