Translator Disclaimer
3 October 2006 On image fusion and segmentation
Author Affiliations +
While the increase in spatial resolution for digital images has been hailed as a significant progress, methods for their automated analyses (i.e. land cover mapping, change analysis, GIS integration) are still in the process of being developed. Object (or segment) based preprocessing techniques seem to be an adequate methodology because inter-class variances can be minimized and the image interpretation techniques of the human eye be mimicked. A number of papers has proven the validity of an segment based image analysis for automated processing, however, the question of appropriate data fusion techniques within this context has hardly been addressed. In this paper, we will investigate techniques for the combination of image fusion and segment based image analysis. The examples include (i) color preserving iconic fusion with subsequent segmentation and classification; (ii) 'cookie cutter' approach for the integration of high resolution RGB and low resolution hyperspectral image data for urban class material detection; and (iii) decision based integration of panchromatic high resolution data with multispectral images for the identification of settlement areas. We will show that the combination of segment based image analysis and fusion techniques at iconic, feature and decision level does indeed improve the final analysis and can be seen as a first step to a an automated result driven processing line. It has to be noted that there is no general theory for segment based image fusion although the feature level fusion seems to be the most promising path for a combination of the two processing paradigms.
© (2006) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Manfred Ehlers "On image fusion and segmentation", Proc. SPIE 6366, Remote Sensing for Environmental Monitoring, GIS Applications, and Geology VI, 63660V (3 October 2006);

Back to Top