Please use this identifier to cite or link to this item:
|Title:||Hybrid framework to image segmentation|
|Authors:||Monteiro, Fernando C.|
|Citation:||Monteiro, Fernando C. (2009) - Hybrid Framework to Image Segmentation. IN Neural Information Processing. Berlin: Springer-Verlag. ISBN 978-3-642-10682-8. p.657-666.|
|Series/Report no.:||Lectures Notes in Computer Science;5864|
|Abstract:||This paper proposes a new hybrid framework to image segmentation which combines edge- and region-based information with spectral techniques through the morphological algorithm of watersheds. The image is represented by a weighted undirected graph, whose nodes correspond to over-segmented regions (atomic regions), instead of pixels, that decreases the complexity of the overall algorithm. In addition, the link weights between the nodes are calculated through the intensity similarities combined with the intervening contours information among atomic regions. We outline a procedure for algorithm evaluation through the comparison with some of the most popular segmentation algorithms: the mean-shift-based algorithm, a multiscale graph based segmentation method, and JSEG method for multiscale segmentation of colour and texture. Experiments on the Berkeley segmentation database indicate that the proposed segmentation framework yields better segmentation results due to its region-based representation.|
|Appears in Collections:||DE - Capítulos de Livros|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.