Content-based query system in image libraries by using color features

ABSTRACT Information retrieval is subject to research and study of information systems for years. Especially, the fast development of Internet enriched text based information presentations, by image, video and voice facilities. For this reason, information searching by only text based methods may be inadequate. Complex multimedia data types, which cannot be represented as text only, requires content based retrieval. Nowadays, it is still common to access databases using text based search, i.e. the use of keywords, to retrieve desired images. In this kind of systems, keywords representing the features such as the subject of the image and objects in its contents are kept in the system for each image. By submitting the desired features via keywords to the system, user can retrieve the images linked to these keywords. However, as the perception of an image may differ from one person to other, it is not an easy job to choose the appropriate keywords to represent an image. In this study, a system described as content based image retrieval using color features is presented. Within the developed system, all images in the database are quantized in different color levels in RGB, HSV and CLE. XYZ color spaces using Median-Cut and K-Means algorithms. Images are represented with specific number of colors obtained; the usage rates and the location of the quantized colors are added to the color features. Query results are obtained by involving a two phased and fast matching algorithm that uses the color features. The similarity of the results to the query image is calculated as a ratio, based on color similarity The developed system is used to perform fast and sensitive queries on image libraries using color features Keywords: Content-Based Image Retrieval, Color Feature, Color Spaces, Quantization, Matching