Image retrieval system based on multi feature extraction and its performance assessment

Abstract

The main problem in the Content based image retrieval is “semantic gap”. A human match two images related to its semantics, while a retrieval system do the same process based on comparison of feature vectors corresponding to visual image features. The problem in retieval of images is identifying its unique features. This article proposes three different methods to retrieve images based on the unique features of an image and also evaluates the retrieval system performance. The first approach called Surrounding information retrieval (SIR) extracts the features related to the similarities of the neighborhood intensity values. The second approach Minimum edge retrieval (MER) identifies the minimum intensity value of each block. The third approach integrated feature retrieval (IFR) combines the properties of feature extraction from SIR and MER. The extracted unique features are stored in a feature dataset and the similar images are retrieved by comparing the dataset using distance measure. The performance of retrieval system is calculated in terms of its recall and precision. The precision and recall values are superior than the existing methods. The method IFR with multi feature extraction shows good in retrieval accuracy compared with other methods.