Comparison between popular image fusion approaches with the consideration of their influence on later segmentation outcomes

Image Fusion is the basic way to complement the panchromatic bands which occupies precise spatial resolution with the multi-spectral bands which contains abundant spectral details. By the courtesy of rapid development of remote sensing techniques, to the same area, there are lots of images with different spatial and spectral resolutions, which boost the necessity of image fusion. There have been various approaches for image fusion available now, most of which emphasize only on the visual effect while ignoring its fusion consequence to the later image analysis processes (such as segmentation, classification, etc). The paper intends to propose some suggestions on how to choose the fusion procedures, based on the comparison of popular image fusion methodologies applied to Quickbird image with 0.6m spatial resolution. The multi-resolution segmentation approach provided by Definiens´ eCognition has been used in the final segmentation stage to ensure the optimal results).