Abstract

Content-based image retrieval (CBIR) is a new but widelyadopted method for finding images from vast and unannotated image databases. In CBIR images are indexed on the basis of low-level features, such as color, texture, and shape, that can automatically be derived from the visual content of the images. The operation of a CBIR system can be seen as a series of more or less independent processing stages. As there exists multiple choices for each of these stages, a multitude of CBIR systems can be implemented by combining a set of common building blocks. In this paper, we present a comparison of different techniques for three consecutive stages of a CBIR system. These include: (1) the initial per feature selection of considered images, (2) the combination of the lists of selected images, and (3) the final selection of images based on all available features simultaneously. The results of the performed experiments show that CBIR systems can be implemented using consecutive stages where, at each stage, a number of parallel techniques can be provided. 1.

Citations

976	An algorithm for vector quantizer design - Linde, Buzo, et al. - 1980
314	Similarity of Color Images - Stricker, Orengo - 1995
146	Self-Organizing Maps, volume 30 of Springer Series in Information Sciences - Kohonen - 2001
78	Self-Organising Hierarchical Feature Maps - Koikkalainen, Oja - 1990
43	PicSOM— content-based image retrieval with self-organizing maps - Laaksonen, Koskela, et al. - 2000
34	K.S.Fu,“Image query by Pictorial example - Chang - 1980
33	Self-organizing maps as a relevance feedback technique in content-based image retrieval - Laaksonen, Koskela, et al. - 2001
28	Introduction to Modern Information Retrieval. Computer Series - Salton, McGill - 1983
25	Modeling user subjectivity in image libraries - Picard, Minka, et al. - 1996
19	Statistical shape features in content-based image retrieval - Brandt, Laaksonen, et al. - 2000