E.S. Hore and S. Ray (Australia)
content-based image retrieval, multidimensional feature combining, image processing
In this paper we present a new algorithm for combining multiple features in content-based image retrieval. The pro posed algorithm allows for a number of features to be se lected by the user and employed in a single retrieval ses sion. In addition, the technique also incorporates automatic feature weighting calculations to determine which features provide the greatest discriminatory power given a query image and images in the database. The user also has the op tion of supplying their own feature weights in order to give more preference to a particular feature. Currently the algo rithm has been tested in combining both colour and shape features with colour being represented in the 24 bit RGB space and shape as moment invariants, centroid, area and circularity.
Important Links:
Go Back
