A Study Of Image Fusion Method Based On Multi-scale Theory

In this paper, we studied the multi-scale image fusion theory and its application. The research contents are as follows:①In this paper, the application of image fusion in medical image, remot image infrared, visible light image and multi-focus image fusion are introduced. Also the image fusion performance of different Multi-scale analysis tools are compared. The image fusion theorys of wavelet transform are described in briefly. And some simple fusion experiments of wavelet transform are made. Although the multi-scale wavelet transform has obvious advantages, it is easily leads to phase deviation and affects the image fusion quality because of the lack of translation invariance;②The spatial frequency ratio which could reflect regional similarity among images was defined. Based on lifting wavelet transform, the nonsampled lifting wavelet transform with the trait of shift invariance was obtained by cancelling the odd and even split process of lifting wavelet transform. In the choice of low-frequency sub-band coefficients, a scheme of the combination of weighting and selection based on spatial frequency ratio is presented; And the high-frequency was fused by weighting fusion algorithm based on edge. The experiment upon multi-focus images are done. The experimental results show that the proposed image fusion algorithm compared with traditional wavelet methods have obvious superiority;③A novel image fusion algorithm is proposed based on nonsubsampled Contourlet transform with the region contrast and spatial frequency ratio as fusion rules limit. For the fusion of visible and infrared image of the same scene, a weighted average scheme based on the physical features of images is proposed for choosing the low-frequency coefficients. With the high-frequency coefficients, a selection principle is proposed based on region contrast and spatial frequency ratio. The experimental results show that the method compared with the traditional simple fusion method and the regional energy fusion rule can be better in visual quality and objective evaluation criteria;④However, in the process of decomposition, nonsubsampled Contourlet transform uses non-sampling laplace pyramid forms, which perhaps leads to appearance of spectrum aliasing, and error of fusion. In order to overcome the drawbacks of pyramid decompose of NSCT, combined with the excellent performance of Non-sampling lifting wavelet, a novel image fusion method is proposed. In this method obtains the nonsubsampled lifting wavelet transform which has the trait of shift invariance. the pyramid decompose of nonsubsampled is taken place by the newly obtained transform. Unifying this new transform with a certain fusion rule, a novel algorithm of image fusion based on the nonsubsampled lifting wavelet-Contourlet transform is obtained. Experimental results show that the new algorithm, compared with the nonsubsampled Contourlet transform method, can extract more useful information from the source image and pour the information into the fusion image, thus obtain a better performed fusion image.