Bilkent University
Department of Computer Engineering


Morphological color image analysis applied to content-based image description, annotation and retrieval


Dr. Erhan Abdullah

Mathematical morphology (MM) is a powerful image analysis framework, nowadays fully developed for both binary and greyscale images. Its popularity is mainly due to its rigorous mathematical foundation as well as its inherent ability to exploit the spatial relationships of pixels. It provides a rich set of tools able to perform from the simplest to the most demanding tasks: noise reduction, edge detection, segmentation, texture and shape analysis etc. This talk concentrates mainly on the extension of mathematical morphology to color images, an issue with relatively simple requirements, yet hardly straightforward. The resulting morphological operators are then applied to the problem of content-based image retrieval; a popular field, focusing on the description and retrieval of visual data, based not on text-tags, but directly on the visual content. More precisely, it is shown that the extension of MM to color and more generally multivariate data, requires a vector ordering, a comparative study of which is presented. We further introduce several variations to lexicographical ordering that take into account the properties of human color vision. Whereas in the second part of this talk, morphology based color and texture descriptors, based on the aforementioned notions, are presented. Bio: Erchan Aptoula received the B.Sc. degree in Computer Engineering from Galatasaray University, Istanbul in 2004. He further holds M.Sc. and Ph.D. degrees in Computer Science from Strasbourg University, Strasbourg, obtained respectively in 2005 and 2008. His research interests include color and multispectral image processing, multivalued mathematical morphology, pattern recognition and multimedia indexing.


DATE: 14 December, 2009, Monday @ 13:40