Application of spectral mixture analysis and image fusion techniques for image sharpening

Abstract

Image fusion is used to merge images of different Spatial and spectral resolutions to create a high spatial resolution multispectral conibination. High spectral resolution allows identification of materials in the scene, while high spatial resolution locates those materials. Spectral mixing is an algorithm that estimates the percentage of each material (called endmembers) within each low resolution multi/hyperspectral pixel. In this article we extend the spectral mixing (unmixing) approach to use one or more higher resolution sharpening images, locating the end-members to higher spatial accuracy. The approach starts with conventional unmixing to generate fraction images. Constrained optimization techniques spatially locate the endmembers to high resolution. Synthetic image generation (SIG) tools are used to generate test images. SIG controls all the image parameters, making it easier to analyze algorithm performance. The results show spectral mixing material maps can be successfully sharpened increasing the effective resolution of the hyperspectral images. Accurate fusion algorithms integrate spectral and spatial information into a single image presenting the most information to an analyst.