الفهرس 
chapter 1Only 14 pages are availabe for public view
 |  | from | 110 |  |  |
| | | from | 110 | | |
Abstract
Content Based Image Retrieval (CBIR) has become a challenging research area, Because of advances in internet and image databases. In CBIR, using visual contents of an image which are mainly color, texture and shape, images are automatically indexed.
These visual features are useful in characterizing an image [2]. Because of these visual features cannot capture any semantic information of an image, the recent trend in CBIR research is to develop efficient features for simpler characterization of images and related matching techniques so that it can handle real life images.
Neutrosophic Sets (NS) theoretic approach may be found suitable in CBIR system because the users are interested in results according to similarity (closeness) rather than equality (exactness). Due to the inherent advantages of this approach of handling uncertainties, NS theoretic model has become appropriate for applications where there is a possibility of incompleteness or perturbation of data.
An image is represented by a set of segmented regions, each of which is represented by a set of Neutrosophic features (texture, shape, color). Similarity between two images is defined as the similarity between the Neutrosophic features and quantified by the similarity measure.
In our proposed technique, we propose a two-phases Content-Based Image Retrieval System (CBIR) for images embedded in the Neutrosophic Domain (ND) which divided an image into three components which are membership degree (T), indeterminacy (I) and non-membership degree (F). Then, constructed their vectors.
In the first phase, we extract texture, shape and color features to represent the content of each image in the training database.
In the second phase, a similarity measurement is used to determine the distance between the image under consideration (query image), and each image in the training database, based in their feature vectors constructed in the first phase using Neutrosophic Euclidean Distance. Hence, the N most similar images are retrieved.