A solution to the correspondence problem in multi-view imagery


Spreeuwers, L.J. and Schutte, K. and Houkes, Z. (1995) A solution to the correspondence problem in multi-view imagery. In: First Annual Conference of the Advanced School for Computing and Imaging, ASCI '95, May 16-18, 1995, Heijen, The Netherlands.

Abstract:This paper concerns the analysis of multi-view imagery in order to obtain a scene description and, specifically, the correspondence problem that occurs in this analysis. The required scene description in this case consists of the classes of the objects present in the scene and their parameters: position, size and orientation. The images are aerial photographs and the objects in the images are man-made objects, such as buildings, roads etc. The goal of the complete system is the fully automatic analysis of aerial photographs of urban areas. The output of the system is a scene description that can be used to generate or update a GIS (Geo Information System). Up to now a system has been developed that performs this analysis on a single image [2, 3]. The advantage of using multi-view imagery compared to using single images is that (partly) occluded buildings may still be recognized, because they can be more clearly visible in other images acquired from a different viewpoint. However, the use of multi-view imagery complicates the analysis, because the objects in the different images have to be corresponded to each other. In this paper a solution to this correspondence problem is presented on object hypothesis level. First the image analysis system for single images is described in short. Then the system is extended for multiple images and the method for corresponding object hypotheses is presented. Finally, experiments and conclusions are given.
Item Type:Conference or Workshop Item
Copyright:© 1995 Advanced School for Computing and Imaging
Electrical Engineering, Mathematics and Computer Science (EEMCS)
Research Group:
Link to this item:http://purl.utwente.nl/publications/16580
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page

Metis ID: 113465