The Study Of Multispectral Image Fusion Based On Six Channel Non-separable Wavelets

Image fusion is the use of a certain algorithm to synthesize a number of images into an image,so that the synthetic image has multiple features.An important branch of image fusion is multispectral image fusion,it synthesizes multispectral(MS)image which focuses on spectral information and panchromatic(PAN)image which focuses on resolution into a new multispectral image which will be good at retaining the spectral and spatial details of the original image at the same time.The multi spectral image fusion technology can be good at solving the problem of multi spectral image with low spatial resolution,which is caused by too much emphasis on the spectral information of objects when shooting.In the field of multispectral image fusion,in order to make the fusion image best preserve the spectral and spatial resolution information of the original image,researchers have proposed many good methods of image fusion such as the IHS transform fusion method,the PCA transform fusion method,the DWT transform fusion method and so on.IHS and wavelet transform are the most representative methods,and many of the following methods are developed on the basis of them.After using The IHS transform method,the fusion image can preserve the spatial details information well.The DWT method is to apply tensor-product wavelet transform to image fusion,the fusion image has blocking artifacts which is cased by using tensor product wavelets without time invariance.The non-separable wavelets have good properties.They can have tight support,symmetry and orthogonality at the same time,and they have the characteristics of isotropy,which make up for the loss of direction of the tensor product wavelet.Therefore,The main research work in this paper is as follows:A construction method of two dimensional six channel non-separable wavelet filter banks is proposed.Constructing some filter banks with compact support,symmetry and orthogonality by using filter construction method which does not depend on the dilation matrix.A multispectral image fusion method based on IHS transform and six channel non-separable wavelet transform is proposed.The implementation steps are as follows:first,transforming multispectral image into IHS space.Second,the multi-scale decompositions of the I and the panchromatic image are performed by using filter banks of six channel non-separable wavelets.Third,the high frequency and low frequency subgraphs are fused respectively according to the characteristics of the decomposed subgraphs.Then,refactoring these fusion subgraphs by using corresponding refactoring filter banks to get the fusion multispectral images.Finally,the performance indexes of the fused image are evaluated and analyzed by visual observation and objective indexes.The experiment results show that the proposed method has good visual effect.The fusion performance outperforms the fusion method based on DWT,IHS-DWT,midway histogram equalization or Guided Filter and Sparse Representation in preserving spectral quality and high spatial resolution information.The spectral information of the fusion result image of the traditional IHS method is distorted.Compared with the traditional IHS method,the proposed method can preserve good the spectral information of the original image.At the same time,compared with the above five methods,the proposed method can preserve good edge information of the original images.