A fusion evaluation approach with region relating objective function for multispectral image sharpening

An approach for evaluating the spatial quality of fused image products is proposed. The method is derived from the imaging process of a step-edge and considers the actual fusion objective instead of relying on subjective assessments. For this purpose, both the panchromatic and multispectral scenes are decomposed into regions according to the spatial characteristics. Then the spatial quality of the fused product is evaluated based on the differences between the objectively expected and the actual resulting region associations in the fusion product. The framework was tested on a number of commonly used wavelet-based fusion algorithms and showed that a reliable quality estimate in terms of spatial information can be derived for various contents and different ratios of spatial resolutions.