Research For The Fusion Of Infrared And Visible Image Based On Multiscale Analysis

Fusion of infrared and visible image becomes more and more popular in the field ofmulti-sensor image fusion. In recent years, the ideas of multiscale geometric analysis havebeen applied to image fusion successfully and become the new concerns. How to constructa fusion rule based on the multiscale geometric analysis, that is accurate and dependable,becomes an urgent question to study. In view of this, we research on the following aspectsin this dissertation.Firstly, the priori information such as the imaging mechanism of infrared and visiblesensors and the imaging characteristic of the source images has been deeply analyzed, andan image fusion algorithm that adapts to the source images has been proposed based onthe contourlet transform. First of all, two original images could be decomposed by thecontourlet transform. And the low frequency image of fusion image can be obtainedadaptively by calculating the local salience factor, besides in connection with thedifference of physical peculiarity between infrared and visible image, an improvedweighted average algorithm has been proposed. The principle of the high frequency imageis the way of combining the directional bandlimited contrast and directional vectorstandard deviation. Lastly, the fusion image could been obtained by performing theinverse contourlet transform. Experimental results demonstrate the rationality of thealgorithm.Secondly, a fusion algorithm of infrared and visible image has been studied based onthe NSCT and PCNN. Compared with contourlet transform, NSCT is shift-invariant,redundant and more suitable for image fusion. Firstly of all, two original images could bedecomposed into a low frequency image and a series of high frequency images of differentdirections separately by NSCT. Then the fusion principle of the low frequency image isbased on the traditional weighted average. The fusion principle of the high frequency isbased on the PCNN, and the algorithm of improved laplacian energy regards as the linkintensity has been studied. Lastly, the fusion image could been obtained by performing theinverse NSCT. Experimental results demonstrate the feasibility of the algorithm. Finally, when the NSCT fusion method is applied, the effects of differentdecomposition levels on image fusion quality are studied. And the experimental resultsdemonstrate the most reasonable decomposition level.