PUB - Publikationen an der Universität Bielefeld

Both contour and region segmentation have their own advantages and drawbacks. In this work we propose to combine contour approximation and contour-based groupings with region segmentation to enhance both contour- and region-based interpretation of the image data. Contour segments and regions are matched based on the distance between contours and region boundary. This information is used to hypothesise alternative regions and to generate additional closed contours within the contour-based grouping process. Initial results for real images are presented.