Fusion of Panchromatic and Multispectral Images Using Non Subsampled Contourlet Transform and FFT Based Spectral Histogram (RESEARCH NOTE)

Abstract

Image fusion is a method for obtaining a highly informative image by merging the relative information of an object obtained from two or more image sources of the same scene. The satellite cameras give a single band panchromatic (PAN) image with high spatial information and multispectral (MS) image with more spectral information. The problem exists today is either PAN or MS image is available from satellite camera. In many remote sensing applications, there is a need for enhancement of the MS image with more spatial resolution for further analysis. In this work, a new fusion technique is proposed in order to get a combined image by integrating PAN image and MS image. PAN and MS images are decomposed using non subsampled contourlet transform (NSCT). Then, the fourth order correlation coefficient (FOCC) is found between the low frequency components of both images. If FOCC is greater than an optimum threshold value, then the corresponding higher frequency components of the PAN image are injected into the respective locations of the MS image. Otherwise MS image components are retained. For calculating the optimum threshold value, n number of test fused images are formed by using n arbitrary threshold values. FOCC between the PAN image and the test fused images are plotted for various threshold values. Similarly, hamming distance of FFT based spectral histogram curves between the MS image and the above test fused images are plotted. The point of intersection of these two curves gives the optimum threshold value. The fused image is obtained using inverse NSCT. The evaluation shows that the proposed work gives better improved results than the other existing methods.