Research On Multi-sensor Image Fusion Algorithm

Image fusion is a technique to combine information from multiple imaging of the same scene,by overall treatment,the fusion image is more clearly and reliable than any single sensor generated.Because of these merits,multi-focus is widely used in these areas such as machine vision,object recognition,the remote sensing image classification and 3d image reconstruction.In this paper,two different types of image fusion are proposed: different-focus image fusion based on multi-level scale blocks(image fusion in spatial domain),color infrared and visible light image fusion based on wavelet transform(image fusion in transform domain).This dissertation laid emphasis on the following points:1.Based on the idea of iteration,a simple,effective and easy to implement multi-focus image method is proposed.Firstly,the two images that need to be processed are Gaussian convolution,and the blurred images are obtained.Then,between the original image and the blurred images are subtract from the different blocks at different levels,and the area decision of image blocks at different levels are obtained.And then in order to get the subjection of image blocks according to the principle of consistency by the area decision of image blocks.As for the area partition map is obtained by using the results of the iteration at different levels of the region division graph.Finally through the method of coefficient weighting get the fused image.The experimental results show that the image quality obtained by this method is superior to the traditional fusion method.2.Through the fundamental research on wavelet,based on good understanding of the fundamental theory and method of image,a new wavelet improved method based on region is presented.This method performs the fusion process finely for both high frequency and low frequency parts of source images.And adopt different fusion rules according to image imaging characteristic.Experimental results show that the proposed method can merge information from source image adequately and improve abilities of information analysis and feature extraction.