A zero saturation distortion image fusion method based on the GCOS framework

Image fusion, widely used in remote sensing, is an efficient method to combine the high spatial information in panchromatic (Pan) image with the essential spectral information in mutispectral (MS) images. In most fusion methods, distortion of spectral information or spatial information commonly exists in the resulting image. The component substitution (COS) fusion methods always merge Pan and MS images with distiortion in spectral information. To overcome this shortcoming, a fusion method based on a general COS (GCOS) framework is proposed in this paper. Landsat TM and SPOT images are used in several experiments to test our fusion algorithm. Some classical indices of images, such as entropy, correlation coefficient, the relative difference and universal image quality index et al, are also used here to evaluate the ability of preserving spectral and spatial information. And the resulting of several methods including IHS, PCA, BT, Gram-Smiths (GS) and SFIM are listed to compare effects of different fusion methods. Experiments show that the proposed method provided satisfied results with richer information than the other methods in both spatial and spectral domains.