An Effective Content Based Image Retrieval System Based on Global Representation and Multi-Level Searching

Retrieving relevant images from a diversified collection using visual queries as search argument is a challenging and important open problem. In this paper the authors present the design and implementation of a simple yet effective Content-Based Image Retrieval (CBIR) system. It uses the colour, texture and shape features. The searching is multi-level with three main consequent searching steps. This proposed system is unique as it considers one feature at each step and uses the results of the prior step as the input for the next step in multi-level manner whereas in past methods all the features are fused at once for the single-level search of a typical CBIR system. The proposed approach is simple and easy to adopt. The retrieval quality of the proposed approach is evaluated using two benchmark datasets for image classification. The proposed system shows good results in terms of improvement in retrieval quality, in comparison with the literature.