Multi-focus Image Fusion Methods Based On The Comibination Of The Filter Bank And The Way Of Sampling

Multi-sensor image fusion is a kind of image information fusion.In this way,different images of the same target or scenarios obtained by using different sensors,or different images obtained by using same sensors in different imaging modalities,or different images obtained in different imaging time, are fused into one result image. The result image can reflect the information of the multiple original images, and describe comprehensively the target of some scene, which is easier for computer processing and more suitable for visual perception. Image fusion is a kind of information fusion researching images. Information fusion is a multi-level processing of testing, combining and estimating the multiple source information to achieve accurate state estimation, identity estimation of situation assessment and complete and timely threat assessment. It can obtain the description of the complete information for target and scene maximum limitedly. Image fusion is an emerging discipline. It combines sensor, image processing, signal processing and artificial intelligence, etc. At the same time image fusion become the research focus in the nowadays because of its wide range of applications, its application areas include:medical image, target recognition, machine vision, robot, remote sensing, complicated intelligent control system, and other areas of the civilian and military fields.There are many methods for image fusion, in order to know the advantages and disadvantages of each image fusion method results more intuitively and conveniently. First,this article studied the basic theory and method about the filter bank structure for two channels and four channels of non-separable wavelets, there are some corresponding examples which are given, and applied to image fusion experiments.. Secondly,This article introduced two different sampling ways:quincunx sampling and the interlacing sampling. According to the different dilation matrix, Quincunx sampling can be used for the decomposition algorithm in two channels and four non-separable wavelet. The interlacing sampleing is a way of interlacing in the wavelet transform,so its operation is simple. Based on combining different filter groups and different sampling ways, four new wavelet decomposition algorithms are presented, and applied to image fusion.Innovation of this paper is to propose the image fusion based on the four channels quincunx sampling non-separable wavelet. And I compare it with the image fusion based on two channels quincunx sampling non-separable wavelet, four channels interlacing sampling non-separable wavelet, four channels interlacing sampling separable wavelet. The four groups of fusion methods are about multi-focus image. Image fusion method is to use the weighted average operator for low frequency coefficient fusion, using the weighted average operator by means of the criterion of absolute value from big to fusion of high frequency coefficient of each resolution, then to reconstruct the fusion coefficients, the fusion image can be calculated out. Based on multiple sets of experimental results are analyzed and compared, the image fusion results gained by using four channels quincunx sampling integral wavelet fusion method can obtain more details than that by using two channels quincunx sampling integral wavelet fusion method; the fusion result image gained by using four channels quincunx sampling integral wavelet fusion method is more close to the standard image and its fusion effect is better than that by using four channels interlacing sampling non-separable wavelet method or channels interlacing sampling separable wavelet method.