Research On Image Fusion Method Based On Image Segmentation And Multiresolution Analysis

Image fusion refers to the process of collecting complementary or redundantinformation obtained by multiple image sensors. The theme is to fuse several samescene multiple images working in different bands into a new image, making the newimage be more suitable for human visual perception or meet the need in the process ofimage segmentation, feature extraction and target recognition. This paper mainlyresearches the infrared and visible image fusion.This paper describes the basic principle of image fusion based on multiresolutionanalysis, mainly including Laplacian pyramid transform, Wavelet transform, Stationarywavelet transform and Non-subsampled contourlet transform. And the quality ofdifferent transforms is analyzed.A novel image fusion framework utilizing image segmentation and multiresolutionanalysis based on traditional multiresolution image fusion framework is researched inthis paper. Firstly, an image segmentation method is applied to segment the infraredimage into target and background part. Then a kind of multiresolution analysis isapplied to fuse the background part of infrared image and visible light image. Finally,the fused background image and the target part of infrared image is fused to the finalfused image by a fusion criterion.An image fusion method utilizing image segmentation and Stationary wavelettransform (SWT) is researched. The infrared image is firstly segmented into target andbackground part utilizing Otsu combined with Canny edge detection. Then a fusioncriterion of taking bigger neighborhood spatial frequency is applied in the fusion ofinfrared background image and visible light image. The final fused image is obtained byfusing the background fused image and the target part of infrared image based onweighted fusion criterion. The test results show that the fusion method can obtain betterresults both in objective and subjective standards.An image fusion method utilizing entropy of the histogram and Stationary wavelettransform(SWT) is researched in this paper. The entropy of the histogram algorithmcombined with Sobel edge detection is applied to segment the infrared image into targetand background part. Then a weight fusion criterion based on gaussian fuzzy logic isapplied in the low frequency sub-band of infrared background image and visible lightimage decomposed by SWT. The final fused image is obtained by fusing the background fused image and the target part of infrared image utilizing weighted sumfusion criterion. The test results show that the image subjective visual effect is improvedeffectively by the fusion method, and the objective evaluation factors are also increasedprominently.