Abstract

We present new insights on the relations between a recently introduced probabilistic formulation of the content-based retrieval problem and standard solutions. New experimental results are presented, providing evidence that probabilistic retrieval has superior performance. Finally, a unified representation for texture and color is introduced. 1 Introduction The problem of retrieving images or video from a database is naturally formulated as a problem of pattern recognition. Given a representation (or feature) space F for the entries in the database, the design of a retrieval system consists of finding a map g : F ! M = f1; : : : ; Kg x ! y from F to the set M of classes identified as useful for the retrieval operation. K, the cardinality of M , can be as large as the number of items in the database (in which case each item is a class by itself), or smaller. If the goal of the retrieval system is to minimize the probability of error, i.e. P (g(x) 6= y), it is well known that the opt...

Citations

6234	Maximum likelihood from incomplete data via the EM algorithm - Dempster, Laird, et al. - 1977
1123	Color indexing - Swain, Ballard - 1991
846	Visual learning and recognition of 3-d objects from appearance - Murase, Nayar - 1995
801	A probabilistic theory of pattern recognition - Devroye, Györfi, et al. - 1996
542	W: Texture features for browsing and retrieval of image data - Manjunath, Ma - 1996
170	Texture Classification and Segmentation Using Multiresolution Simultaneous Autoregressive Models - Mao, Jain - 1992
103	directionality and randomness: Wold features for image modelling and retrieval - Liu, Picard
51	Real-time recognition with the entire Brodatz texture database - Picard, Kabir, et al. - 1993
40	Benchmarking of image features for content-based retrieval - Ma, Zhang - 1998
20	A Bayesian framework for semantic classification of outdoor vacation images - Vailaya, Figueiredo, et al. - 1999
13	Library-based coding: A representation for efficient video compression and retrieval - Vasconcelos, Lippman - 1997
11	Embedded mixture modeling for efficient probabilistic content-based indexing and retrieval - Vasconcelos, Lippman - 1998