Multispectral And Panchromatic Image Fusion Algorithm Based On IHS Transform

With the development of remote sensing satellite technology, remote sensing images play an important role in daily life and military application. Multispectral images (MS) with higher spectral resolution and spectral information is abundant. However, the spatial resolution is lower due to the physical device restrict. Contrary, panchromatic images(PAN) have a higher spatial resolution and distinct spatial details. But the spectral resolution of PAN is lower. The fusion of multispectral image and panchromatic image is to inject spatial detail extracted from panchromatic image into multispectral image, and synthesize a higher spatial and spectral resolution multispectral image. Multispectral images with higher spatial resolution are better for terrain classification, target recognition, and other processes subsequently.This thesis mainly analyzes the multispectral images and panchromatic images, of which the spatial resolution is 4:1. The main research contents and innovations of this thesis include the followed aspects:(1) We analyze the image fusion algorithm based on IHS transform. One of the major reasons for spectral distortion is the signification difference between substituted component and substitution component. This thesis proposes the improved ideas. Then we discuss pansharpening algorithm based on local correlation of the remote sensing image. The algorithm has the advantage of high spectral fidelity and the disadvantages of high time complexity and complex fusion rule. As a result, an image fusion algorithm based on IHS transform and NSCT is designed. Firstly, the algorithm resamples MS image using bicubic interpolation, transforms it into IHS color space, and decomposes intensity I and panchromatic image P. For preserving spectral information of multispectral image and detail information of panchromatic image, low frequency coefficients of intensity I are selected as low frequency coefficients of fusion image. High frequency coefficients of panchromatic image P are selected as high frequency coefficients of fusion image. The algorithm has a simple fusion rule and combines the advantage of image fusion’s fast implementation based on IHS transform and the advantage of image fusion’s high fidelity based on NSCT. Experimental results demonstrate that fusion images have better spectral fidelity and clearer texture.(2) We analyze the shortcoming of interpolating image using bicubic interpolation. To solve the problem, the algorithm interpolates intensity I, which is achieved by space transformation by spare coding with two steps to achieve abundant spatial detail. The algorithm designs rules of fusion based on edge protection, which takes the pixel and the local window into account in order to preserve textures and edges of panchromatic image. A new intensity S is obtained by inverses transform. Then the scheme extracts edge information from panchromatic image P using Canny operator. Pixel values of new intensity’s edge positions corresponding to the edge positions of panchromatic image P is substituted by pixel values of panchromatic image’s edge positions to further protect the edge information of panchromatic image. A final substitute component PI is obtained. Finally, the scheme transforms space transformation into the RGB color space by inverse transformation and obtains the high spatial resolution multispectral image. Experimental results demonstrate the algorithm’s fusion images have high spectral fidelity, clear texture, and better visual perception.