In this paper we introduce a new approach aimed at solving the problem of image retrieval from text queries. We propose to estimate the word relevance of an image using a neighborhood-based estimator. This estimation is obtained by counting the number of word-relevant images among the K-neighborhood of the image. To this end a Bayesian approach is adopted to define such a neighborhood. The local estimations of all the words that form a query are naively combined in order to score the images according to that query. The experiments show that the results are better and faster than the state-of-the-art techniques. A special consideration is done for the computational behaviour and scalability of the proposed approach. © 2010 Springer-Verlag Berlin Heidelberg.
CITATION STYLE
Paredes, R. (2010). Large-scale text to image retrieval using a bayesian K-neighborhood model. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 6218 LNCS, pp. 483–492). https://doi.org/10.1007/978-3-642-14980-1_47
Mendeley helps you to discover research relevant for your work.