Research On High Resolution Remotely Sensed Image Fusion Methods And Evaluation Of Fusion Quality

With the rapid development of remote sensing technology, the research of image fusion between Multi-sensor is brouht to the forefront by scholars increasingly.But the research of fusion with high resolution remote sensing image is less relatively. Using the experimental data what is multi-spectral and panchromatic data of ALOS and SPOT, basing on the pixel level fusion, the article discussed the fusion techniques and methods of high-resolution remote sensing image, and evaluated the fusion results.This paper reviewed the development of remote sensing image fusion, and summarized the present the basic theory of remote sensing image fusion. It also explained the steps and levels of remote sensing image fusion, researched the basic characteristics and principles of each level of fusion and described the fusion algorithm for each level. The paper elaborated the key step in remote sensing image fusion, which included geometric correction, image registration and image histogram matching. It showed the importance of those work for remote sensing image fusion. This made sure that the correction and image registration accuracy could meet the requirements of remote sensing image fusion and the fusion might achieve the best results.And then described the traditional fusion method which included transform fusion of IHS, PCA (principal component analysis) fusion, Brovey fusion as well as Gram-Schmidt transform fusion. In addition, it improved the SFIM algorithm for multi-spectral and panchromatic images of ALOS. Using the complementary characteristics of multi-spectral image and panchromatic image, SFIM Fusion algorithm can modulate the spatial information reasonably and effectively from high-resolution remote sensing image to the multi-spectral image which has been regulated, and it does not change its spectral properties and contrast. In order to make the fused image improving the contrast, containing more texture information, reflecting the surface features boundary information, this article added texture parameter variance to the SFIM Fusion algorithm. The method is that normalizing the texture of the whole lot of image after calculating the texture of the full color image and adding the parameters obtained to SFIM fusion algorithm. It added the regional variance information to wavelet fusion for multispectral and panchromatic images of SPOT. In the wavelet fusion algorithm, the high frequency components contain more brightness mutations in the remote sensing image, which is more detailed information. Image variance can be expressed as surface features of the boundary information. Before fusion of wavelet coefficients, we can calculate the variance by field operations for high-frequency part of wavelet decomposed. The variance calculated can reflect the more detailed boundary information. The fusion algorithm is more optimal than simply using wavelet fusion.Finally, the article summarized the evaluation which integrated the subjective and objective quantitative assessment of the fusion result for the multi-spectral and panchromatic images of ALOS and SPOT. By comparing the improved integration and the traditional fusion algorithm, it showed that the two improved fusion algorithms in this paper were superior to other algorithms for the multi-spectral and panchromatic images of ALOS and SPOT.