Content-Based Image Retrieval System with Combining Color features and Gradient feature

Abstract

According to the content-based image retrieval system, this system has low retrieval accuracy problems and high computational complexity. A novel method for content-based image retrieval system, proposed in this paper based on Color and gradient feature. Color-histogram and Color-moment are used for Color feature and as its gradient feature, edge histogram feature (SRF) is adopted based on the Local Binary Pattern.Besides the use of the above mentioned features, the Principle Component Analysis has been used for decreasing the dimension of features. After performing PCA upon each feature, the weight was assigned to the Color feature and gradient feature, and similarity was also calculated based on normalizing Canberra Distance. Experimental results showed that in the proposed method "retrieval accuracy" is higher than conventional methods.