Feature Selection in Example-Based Image Retrieval Systems.
Vikas Prasad, A. G. J. Faheema, Subrata Rakshit · Indian Conference on Computer Vision, Graphics and Image Processing · 2002
The objective of Content Based Image Retrieval (CBIR) systems is to retrieve images from large datasets based on queries regarding their contents. This paper discusses the problem of selecting features for handling generic queries in Example-Based Image Retrieval, where the queries are given in the form of positive and negative examples. No assumptions are made regarding the nature of images or queries. We investigate several linear time-complexity algorithms which can be used for selecting features optimal for a given query. We test three aspects of the algorithms: their discrimination ability, robustness to sample set sizes and behavior in real world test cases. The results indicate that a hybrid approach may be called for.