Evaluation of Two Applications of Spectral Mixing Models to Image Fusion

摘要

Many applications in remote sensing require merging low-resolution multispectralor hyperspectral images with high-resolution panchromatic images to create high-resolution multispectral or hyperspectral material maps. A number of methods are currently in use to produce such hybrid imagery. Until now, these methods have only been evaluated independently, and have not been compared to one another to determine an optimum method. This research performed a quantitative test of three image fusion procedures. The first method involves first sharpening low-resolution multispectral data using the panchromatic image, to produce a high-resolution multispectral image. This image was then separated into a series of high-resolution images which provided a mapping of materials within the scene. The second method involved first separating the low-resolution multispectral data into a series of material maps using a recently developed adaptive unmixing algorithm. These maps, along with the panchromatic image, were used to produce high-resolution material maps. The final method examined involved creating the low-resolution material maps using traditional image-wide unmixing methods. The resulting images, along with the panchromatic image, were used to produce sharpened material maps. These three image fusion procedures were evaluated for their radiometric and unmixing accuracy. It is hoped that the optimum method identified by this research will enable analysts to more easily and accurately produce high-resolution material maps for various applications.