Improving the Effectiveness of the
Color Coherence Vector

Alaa Al-Hamami1 and Hisham Al-Rashdan2
1Computer Science Department, Amman Arab University for Graduate Studies, Jordan
2Computer Science, Yarmouk University, Jordan

Abstract: This paper was motivated by the desire to improve the effectiveness of retrieving images on the basis of their color content by improving the Color Coherence Vector method. There is a growing demand for efficient and effective image retrieving techniques from the huge multimedia databases on the Internet and from the image libraries. In this paper three methods have been proposed and developed to improve the effectiveness of the Color Coherence Vector  method: The modified Color Coherence Vector based on the number of the color coherence regions, The modified Color Coherence Vector based on the distance of the color coherence regions and The modified Color Coherence Vector based on the angle of the color coherence regions. The proposed methods take advantage of the coherence regions number and location information. The information of the coherence regions number, distance and angle along with the coherence and incoherence pixels amount for each color was presented by using the simple and flexible histogram representation technique. The experiments were carried out on a collection of 1014 color images of which 100 query images were used as stimuli for retrieving similar images from the image collection. The experimental results indicate that the three proposed methods perform better than Color Coherence Vector  in terms of retrieval effectiveness; the developed methods produce better recall-precision curves and produce better ordering values in terms of Eavg (Ratio (average-rank to ideal-average-rank)). These results provide evidence for the importance of the coherence regions number and location information which have been neglected by the original method color coherence vector.

Keywords: Content-based image retrieval, color coherence vector, color-based image retrieval, and color indexing.