The Research Of Image Fusion Algorithm Based On Wavelet Transform And Canny Operator

Image fusion is a hi-tech which combines sensor, image processing,computer, artificial intelligence tech and many other modern technologies. Thistechnology integrate the source image into one image which describe the sceneclearly and comprehensive by using information complementary and temporalcorrelation which collected by sensors on the same target. Image fusion can befound many applications in the field of remote sensing detection, trafficmonitoring, environmental protection,safety navigation, etc.This thesis introduces the idea of edge into image fusion and studies theimage fusion algorithm which based on wavelet transform and the improvedCanny operator. The main contents of this thesis are as follows:The first section briefly introduced the background, significance, conceptof images fusion and some methods used in this field. Some image fusionmethods based on pixel level are mainly discussed. In addition, image fusiontheory based on wavelet transform and image edge detecting theory are detaileddiscussed.The second section focus on image fusion algorithm based on wavelettransform. In the process of image fusion, the different selection of wavelet basisfunction, layer number of wavelet decomposition and fusion rules will affect theeffect of image fusion. In this article, through some experiments to determinethe best decomposition layer selection and optimal selection of wavelet base,and the images after wavelet decomposition, due to its own characteristics of the different frequency components，an unique fusion rule was discussed in-depth,experiment analysis and comparison of results.The third section, in order to can get more information on the details andimprove the disadvantages of traditional image fusion methods, the section setsidea of image edge into the traditional image fusion methods. Focus on theprinciple of Canny operator is analyzed, and for the uncertainty of Gaussianfiltering parameters as well as double threshold values in the process ofextraction, the thesis gets a Canny algorithm based on adaptive threshold ofstatistical filtering.The fourth section, the wavelet transform and the improved Canny operatorcombined with a new image fusion algorithm is put forward. The low frequencypart uses the weighted average method. And the high frequency part of the firstuses Canny operator to extract the high frequency sub image edge, and thenmakes the fusion according to regional energy on the edge, finally the fusedimage is reconstructed. This section choose three kinds of model experimentwhich contains multi-focus images, infrared and visible images and medicalimages. The experimental results show that, whether it is based on spatialfrequency, average gradient and the information entropy of objective evaluationcriteria or subjective evaluation, the proposed selection methods aresuperiority and rationality.