Multi-focus Image Fusion Algorithm Based On W Transform

Image fusion is one of the research hotspots in the field of image processing and analysis.Image fusion technology has been widely used in computer vision,automatic target recognition,imaging navigation and guidance,intelligent machine manufacturing,etc.Image fusion is process of combining relevant information from two or more some source images of the scene obtained by different sensors into a single image which can provide much more visual information than the source images.Multi-focus image fusion is an important approach to obtain a composite image with all objects in the scene in focus.In this paper we take multi-focus images as research object,the multi-scale analysis method as research tool,improving the quality of fused image as the research goal.On the basis of analyzing and summarizing the existing frequency domain based fusion algorithms,a new mathematical tool,the W system of degree k,is introduced in this paper in consideration of the disadvantages of poor noise immunity,high computational complexity and unsatisfactory fusion effect of some existing image fusion algorithms.The W system of degree k is a hybrid function system consisting of piecewise polynomials of degree k.By combining the W transform with Laplacian pyramid,gradient pyramid,Contourlet and non subsampled Contourlet transform,four improved multi-scale image decomposition methods are proposed.On this basis,according to the characteristics of the new decomposition method the corresponding fusion rules are put forward.Accordingly,four improved multi-focus image fusion algorithms are proposed.The main results of this paper are as follows:(1)Based on the W system and Laplacian pyramid,a new image fusion algorithm is proposed.In the new algorithm,source images are decomposed by LP and W systern,and according to the characteristics of its decomposition we use fusion rule of local energy weighted threshold value to fusion.The experimental results show that the new algorithm can eliminate the block effect and pseudo-Gibbs phenomenon and improve the objective evaluation indices.(2)Combining W transform with Contourlet transform,a new multi-scale decomposition algorithm is proposed.According to the characteristics of the new decomposition algorithm,for the low frequency we use fusion rule of texture weighted threshold matching method.Accordingly,a new multi-focus image fusion method called WCT algorithm is proposed.A large number of simulation experiments show that the new fusion algorithm can improve the visual effects of fusion images and evaluation index effectively.(3)Combining W transform with non subsampled Contourlet transform,a new multi-focus image fusion algorithm called WNSCT is proposed,which uses the improved pulse coupled neural network(PCNN)as high frequency fusion rules,in the improved PCNN fusion rules,the information entropy and local visibility are used to replace the fixed value in the classical PCNN as the link strength.The WNSCT algorithm can effectively suppress the artifact information and produce a fusion image with better visual effect.The objective evaluation indices including standard deviation,information entropy,mean gradient and spatial frequency are superior to that of the traditional Contourlet transform,non sampled Contourlet transform and the discrete transform.(4)Based on the evaluation indices of multi-focus image,the qualitative evaluation indices of multi-focus image are summarized,and the standard deviation,information entropy,average gradient and spatial frequency are selected as the evaluation indices to evaluate the effect of the proposed fusion algorithm in order to better evaluate the proposed new algorithm and fusion rules.