Date of Award
CC BY-NC-ND 4.0
Large image databases containing a wide variety of imagery are increasingly more common. Because of their diversity and size, they are often poorly indexed. As a result, new automated techniques which index and search images using visual image properties, such as color and texture, have emerged. Texture is used to charaterize image regions and requires the regions of an image to be identified prior to indexing. Two region extraction methods: subdivision and segmentation are used for this task, and are used to compute direct-match and object-based queries, respectively. Direct-match queries are less costly to compute, and indexing is done without user intervention, resulting in automated content-based image retrieval systems (CBIR). In this thesis, we investigate the use of a new subdivision method, fuzzy image subdivision, to address the undesirable dependence on object position in direct-match texture queries. We implement our approach for querying images on the web and compare the retrieval performance with the current direct-match texture method based on rectangular partitioning, which does not take into account changes in object position. Paper copy at Leddy Library: Theses & Major Papers - Basement, West Bldg. / Call Number: Thesis1998 .I53. Source: Masters Abstracts International, Volume: 39-02, page: 0526. Adviser: Joan Morrissey. Thesis (M.Sc.)--University of Windsor (Canada), 1998.
Ingratta, Donato., "Texture image retrieval using fuzzy image subdivision." (1998). Electronic Theses and Dissertations. 3742.