Content-Based Image Retrieval Using Color and Texture Features Through Ant Colony Optimization

Abstract

Content-based image retrieval (CBIR) is retrieving relevant images from the large image database through visual characteristics. Each image in the database and query image is represented through feature vector derived from color and texture features in the image. These feature vectors are compared for relevance to obtain similar images in CBIR system. Therefore, length of the feature vector is very important in the CBIR system. Higher length of the feature vector increases number of comparison and in turn, increases the computational complexity, whereas lower length of the feature vector reduces comparison and complexity. In this paper, performance of the proposed CBIR system using color and texture feature extraction through histogram and Gabor wavelet transform, respectively, is presented. It is necessary to extract all the features of each image from image database and query images. These features are further presented for ant colony optimization to reduce the length of the feature vector. These final features are used in image retrieval process. Experiment results clearly show that the proposed CBIR system through ant colony optimization algorithm performance is better than other algorithms by 1.8% with respect to precision and recall. Also, the proposed algorithm clearly demonstrates the improvement by 10% on the precision and recall using only color and texture features. One of the biggest advantage and improvement was reduction in retrieval time in comparison with the other algorithms.