Chinese Journal of Computers   Full Text
Title	Clonal Selection Remote Sensing Image Fusion Based on CP and Multiwavelet HMT Models
Authors	JIN Hai-Yan1),2) JIAO Li-Cheng2)
Address	1)(School of Computer Science & Engineering, Xi’an University of Technology, Xi’an 710048) 2)(Institute of Intelligent information processing, Xidian University, Xi’an 710071)
Year	2009
Issue	No.7(1434—1442)
Abstract &   Background	Abstract How to obtain efficient fusion coefficients is the key problem in image fusion processing. In terms of the statistical characteristic of images, CP decomposition and GHM multiwavelets are constructed and using multiwavelet domain HMT models to capture the dependencies of coefficients in this article. Furthermore, the evolution computation idea—Immune clonal selection (ICS) algorithm is introduced to optimize the fusion coefficients for better fusion results. Fusion performance is evaluated through subjective inspection, as well as objective fusion performance measurements. Results clearly demonstrate the superiority of this new approach when compared to conventional wavelets and multiwavelet systems as information entropy (IE) values keep at a high level, and average grads (AG) values increase averagely about 1.3 and 2.3, respectively and standard differences (STD) values increase averagely about 8.0 and 8.8, respectively. Keywords image fusion; immune clonal selection; multiwavelet transform; HMT models; CP decomposition Background Statistical image processing in multiscale transform domain has received widespread attentions in image processing, pattern recognition and computer vision in recent years. In practical remote sensing image fusion applications, the key problem is to obtain efficient fusion coefficients. For 2-D images, the aim of fusion is to extract all the useful information from source images to generate a single image. The fused images should contain integrated information, which is more useful, exact, comprehensive, and reliable for future human of machine perception. At present, most studies are concentrated on the pixel level. The best-known methods are simple fusion method, pyramid-based methods, and wavelet-based methods. In this article, a new approach based on ICS-CPMWHMT method is presented. Fusion performance is evaluated through subjective inspection, as well as objective fusion performance measurements. Results clearly demonstrate the superiority of this new approach when compared to conventional wavelets and multiwavelet systems as information entropy (IE) values keep at a high level, and average grads (AG) values increase about 1.3 and 2.3, respectively and standard differences (STD) values increase about 8.0 and 8.8, respectively. This work has been partially supported by National High Technology Research and Development Program (863 Program) of China under grant No.2001CB309403, Natural Science Foundation of Shaanxi province under grant No.2007F51. This group has been working on image processing including image fusion, image denoising and image segmentation, etc. for many years. So far, they have obtained some achievements in this area. By the end of 2007, the group has published more than twenty papers in the international proceedings and journals on this topic, among which about ten papers are indexed by SCI and more than ten papers by EI.