Color-Texture based Feature Modeling for Content based Image Retrieval

V.G. Ranjith, M.K. Jeyakumar and S. Palanikumar

Content Based Image Retrieval (CBIR) is regarded as the most suitable system for the fast and accurate retrieval of images from the large databases over the interconnected networks. It works by collecting images that are similar to the image that is requested from the tremendous database. The main module of the CBIR is a color and texture feature extractor which is used for extracting the major features. The main scope of the work is to introduce an improved analysis which can collect the most similar images from the vast collection of database. Images which have anisotropic and oscillatory textures are very difficult to analyze and represent. This is the main drawback of the existing CBIR systems. Hence, a new combination of Color Co-occurrence Feature (CCF) and Wave Atom Feature (WAF) extractor is proposed in this work. The different types of multi resolution features used such as Wavelet, Contourlet and Curvelet. The wave atom technique can adapt to any local patterns and can be used to represent images that are anisotropic. The proposed system uses a novel similarity score method to retrieve images from the database. The database used two of the most widely utilized benchmark databases like Corel and MIT visTex. From the experimental results, it shows that that the proposed system has outperformed the traditional method. It has improved both the precision and recall rates, where the average precision rate was 0.925 and recall rate was 0.743.

Volume 11 | 04-Special Issue

Pages: 245-254