Evaluation Criteria for Image Fusion Performance in Different Applications

Image and data fusion is a process dealing with data and information from multiple sources to achieve refined or improved information for decision making. It has wide applications. When images with a similar acquisition time are used, the expected result is to obtain a fused image that retains the spatial resolution from the panchromatic image and color content from the multi-spectral image, to perform land use/cover classification, or to extract geo-spatial information, which can be used in thematic mapping, as well as the establishment and update of geo-spatial databases; when images with different dates are used, the main purpose of image fusion is to detect the changes over a period of time, which can be employed in disaster monitoring, environmental monitoring, land use/cover change detection, etc. Aiming at the two types of applications, this paper discusses and summarizes the evaluation methods for image fusion performance in the level of pixel-based processing and decision-based processing. It also develops a quantitative analysis method for assessment of the pixel-level image fusion for change detection. This method integrates spectral features and spatial texture features which constitute the most important visual content of an image. The main idea of this method is to compare the image similarity between the regions where changed parcels are located and the regions where there is no change using the similarity measure. The practical application examples of the evaluation criteria are also given.