Multi-spectral Remote Sensing Image Registration And Fusion Research

Multispectral image has abundant spectral information which creates favorable conditions for recognizing the earth's border and surface, so it has a wide range of applications as very important image.In the paper, the registering and fusing methods of multi-spectral remote sensing image are discussed.At first, this paper introduces three corner-detection algorithms which apply to the common gray images. They are Harris, SUSAN and CSS corner detection operator.Then the image registration based on SIFT is studied in detail in this paper.four approaches are used in registering of multi-spectral remote sensing image above. The experimental results demonstrate that the extracted feature points of the Harris algorithm are scarce and the extracted feature points of the SUSAN algorithm are insensitive to contrast. So the feature points of matching can hardly finish the image registration.ACSS and SIFT algorithm can extract many feature points in the individual band image.Both methods can be very good to complete image registration to multi-spectral remote sensing image which has little difference in the rayscale and spectral bands. But thematching effect is not good to multi-spectral remote sensing image which has much difference in the rayscale and spectral bands.According to Multi-spectral remote sensing image which is characteristics of the images with different gray level properties and subpixel position shift, this paper proposes a new method based on local gray extremum method. Firstly it is need to find small areas containing gray-extremum in the reference image and save the areas. Determine to also contain local gray extremum in the registration image or not. If there is, we keep the area, or else it is eliminated. Secondly the extremal regions will be surface fitting with polynomial. Lastly, extreme points of the small area are calculated respectively as feature points for registration. In this paper, it is experiment with real and simulated multi-spectral images. The results show that the proposed approach has the advantages of simple algorithm and good precision.On the basis of image registration, this paper also introduces image fusion algorithm. Image fusion being an important application of image registration,this paper studies multi-band image fusion based on wavelet transform and the fusion of Panchromatic and Multi-spectral image based on the traditional wavelet transform,HIS transform and the combination method of HIS and wavelet tranform. The experimental results demonstrate that fusion result is the best to the multi-band image when low frequency coefficient adoptes weighted average and high frequency coefficient applying to Improved Laplacian rules.For the Panchromatic and Multi-spectral image, the fusion result based on HIS and wavelet transform method is better than the other two.