Fusion Algorithm For Multi-focus Images Based On Signal Decomposition And Classifier

As a form of information carrier,images are commonly used in social life.It contains the relevant information of the described object,which is used to constructs the visual thought pattern.Those thought pattern will assist human in learning and culture communication.Due to the difference of source,images that describe same scene contain different information.Normally,this information have various uses of their own,such as medical anatomical imaging and medical functional imaging,remote sensing images with different spectral and spatial resolution,etc.The multi-focus Images we study to fuse is one kind of picture groups that captured by same camera at the same position with different focus.And as the consequence of the limitation of the depth of field,only the objects within the depth of field are sharp,while other objects are blurred.Multi-focus image has widely research value,because of the widespread use of the optical photographic equipment.As the multichannel data collected from the same target image is redundancy and complementarity,we can perform image fusion on those images to improve the resolution and information utilization of original images,as well as to enhance calculation precision and reliability.Multi-focus image fusion is an important branch of image fusion,which has been widely used in vision-related processing tasks,such as image segmentation,edge detection and stereo matching.It's preferred that all the objects in the scene are clear,when it comes to vision-related processing tasks,it will greatly improve the efficiency of image processing,reduce costs and improve quality.In this paper,we first briefly introduce some related technologies and the existing multi-focus image fusion algorithms,and the feasibility of applying signal decomposition mechanism in multi-focus image fusion is analyzed according to the existing algorithms.Then we proposed a composite component analysis algorithm(MMCA)based on morphological component analysis(MCA)via intensive studying in compressed sensing and sparse representation.The MMCA algorithm is applied to the transform domain based multi-focus image fusion algorithm,and the image feature is extracted which is utilized in fusion rules as weight value.This fusion algorithm we proposed is denoted as FWMMCA(Fusion Algorithm Based on Weighed Composite Morphological Component Analysis).The advantages and shortcomings of the proposed algorithm and the existing algorithms are analyzed based on the experimental data in various groups of multi focus images.In order to optimize the proposed algorithm,we analyze the possibility of applying the classifier to the multi-focus image fusion instead of the focus detection algorithm.Then success in employing the support vector machine classifier(SVM)into the fusion algorithm based on MMCA,by designing the sampling mode and the fusion rules.Thereafter,we perform consistency verification on SVM to improve the accuracy of classification,which improve the quality of the fused image as well.And this novel fusion method we proposed is denoted as MMCA-CSVM(Fusion Algorithm Based on Composite Component Analysis and Support Vector Machine with Consistency).Finally,Experimental results demonstrate the superiority of the proposed method in terms of subjective and objective evaluation that including Mutual Information(MI),Peak Signal to Noise Ratio(PSNR),structural similarity for multiscale image(SSIMF),Piella's metrics(Q,QW and QE).