الفهرس 
Introduction
Search processOnly 14 pages are availabe for public view
 |  | from | 121 |  |  |
| | | from | 121 | | |
Abstract
Due to the huge number of images available via the Internet, how to search for the required images from the internet quickly and effectively has become an important research direction in image retrieval systems. Search engines are the most powerful resources for finding images on the rapidly expanding World-Wide Web. These search engines retrieve images according to the meta-data of images or the context in which the images are found (the web pages for example).
The image retrieval efficiency of such engines relies heavily on the matching between the user’s text query and image description not on the internal features of images. Therefore, the user can be overwhelmed by thousands of results retrieved by a search engine, of which few are valuable. To overcome the deficiency of text based image retrieval systems, the content based image retrieval (CBIR) systems are introduced. CBIR systems search for images using image content such as color, shape and texture and require the user to submit a query image, and return images that are approximately similar to the query. This thesis presents an approach for distributed Content based Image Retrieval. The proposed approach is based on agents technology where agents autonomously search for images over the internet, then the image features (color and shape features) are extracted from the images and stored in the database. The extracted features are used as the basis for grouping the images servers that contain similar images categories together.
The proposed approach aims at reducing the network traffic by avoiding the transfer of huge intermediate data through the network. The proposed approach is based on a cluster algorithm by which the similar images at each image server are categorized together into a cluster. The centroids of these clusters are sent to a central directory server. Fortunately, the proposed approach reduces message transfer as there is no need to update the cluster centroids with each new image. Instead, the need for a new calculation of the clusters arises after a significant number of updates.
Different measures are used to evaluate the performance of the proposed approach. The results show that the proposed approach reduces the network traffic and also improves the total time taken to retrieve the query results compared with the other current approaches that used mobile agent in image retrieval.
The results show also that the image retrieval by color feature only retrieves more irrelevant images while the image retrieval by color with shape methods outperforms the result with color feature only.