The Development of Color Based Visual Search Utility

During the past few years, much attention has been paid to manage the overwhelming accumulation of rich digital images. In order to improve the traditional text-based or (Structured-Query-Language) SQL-based databases, researches focused on accessing large image databases by the contents of images, such as colors, shapes, and textures. As a result, several content- based image searching systems or met hods were developed. In this thesis, the issue of color-based image search was addressed with special emphasis on color feature. An introduction to color perception, the theoretical foundations of the human image retrieving process, and the content-based image systems and their uses was presented. Several systems were developed. These systems modelled image data using features such as color, texture and shape. Such features are usually extracted from images and stored into database index. Color is one of the most recognisable features exercised by people for visual distinction. Based on observations on how humans measure the perceptual similarity of images, recent studies concluded that human beings have a limited color perception range. Expediting these conclusions, firstly, perceptual color palettes to be used as the perceptual threshold were defined. Secondly, the color algorithm was developed to interpret natural expressions of content such as 10%, 20%, etc. The database-indexing algorithm designed to be independent to the database. Finally, a binary search algorithm was used to match and display images requested. This approach is unique because it is based on hybrid approach to the color based image search. This developed system can be used for any real-world online database. The system was implemented using Microsoft Visual C++ programming language and HTML. Using 200 images as an experimental database, results of the prototype software demonstrated the achievement of the perceptual concept in image content search