Research On Multilevel Mixed Technology Of Multi-sensor Images Fusion

As the different imaging mechanism of optical image,infrared image and SAR image,having the large different characteristics between images,so fusing three types of images withimage fusion technology could complement advantages and be able to better interpret thescenes information.However, information reflecting the goals of the three types of images isvery different,it is difficult to obtain great fusion effect with conventional methods,so fewresearch currently underway among the three types of image.This article studies the fusion ofthree types of images for the purpose of target recognition with SVT transform, mathematicalmorphology and sparse representation,analyzed the relevance and synergy between pixel leveland feature level fusion algorithm, presented a mixed fusion algorithm between the threetypes of images,established a multi-level mixed color image fusion model with colorinformation processing and gray image colorization method.Research work can be divided into four parts:First,a pixel level and feature level mixed fusion algorithm is presented between infraredand optical images. To fuse images using multi-scale decomposition could overall thedifference information of two source images very well,but it can’t solve the inherent problemof edge region blur and low contrast in infrared image.For this issue,in the beginning thearticle fuses images with SVT transform,then extracts bright and dark detail features from thetwo source images using Top-Hat transform,at last weights features to the initial fusedimage.Experimental results show that this method enhances the edge region, improve thecontrast and obtain better fusion image.Second, a fusion algorithm of SAR and optical images with fast sparse representation onlow-frequency images is proposed.For the disadvantage of target information easily missingand the contrast low in fused image,and the fusion method with sparse representation couldeffectively retain target information of SAR image,so the article fuses low frequency imagesof SAR and optical images using sparse representation.Moreover a new sparse coefficientfusion rules is proposed,and sparse decomposition process is improved to reduce thealgorithm running time.Experimental results demonstrate the effectiveness of the algorithm. Third,a color enhancement algorithm combining color space mapping and colortransferring is proposed to establish multi-level mixed color fusion model between the threetypes of image.For the fused images above,the second fusion is carried out,and then processthe final fused image with false color to improve the ability of scene description and targetrecognition.Fourth,according to the human visual characteristics,a evaluation method of color fusionimage is proposed for target recognition combining subjective and objective indicators.Bycomparison with the source images,the effectiveness of the fusion model is validated.