Content-Based Image Retrieval Hybrid Approach using Artificial Bee Colony and K-means Algorithms

In this paper, a new clustering method is proposed for CBIR system; this method depends on combining ABC and k-means algorithm. Four features are used with the proposed method to retrieve the images. These features are extracted by: color histogram of HSV image and color histogram of opponent image to describe the color, Gabor filters and Ranklet transform for RGB image to describe the texture. The proposed hybrid clustering method is a clustering process for database of each feature using k-means algorithm enhanced by ABC algorithm. The innovation in this approach is that each solution in ABC algorithm represents the centroids of clusters that come out from applying k-means algorithm. The proposed method is applied on Wang dataset (1000 images in 10 classes) and evaluated by comparing the test results of the proposed scheme with another existing method uses same database. The results proved that the proposed method is superior to the existing method in terms of the precision in 6 out of 10 categories of WANG dataset, such that the average of the precisions for all categories is 0.8093.