Content-Based Image Retrieval System Using Feed-Forward Backpropagation Neural Network

Extensive digitization of images, paintings, diagrams and explosion of World Wide Web (www), has made traditional keyword based search for image, an inefficient method for retrieval of required image data. Content-Based Image Retrieval (CBIR) system retrieves the similar images from a large database for a given input query image. Today, we find various methods for implementation of CBIR which uses low-level image features like color, texture and shape. In this paper, a global image properties based CBIR using a feed-forward backpropagation neural network is proposed. At first, the neural network is trained about the features of images in the database. The image features considered here are color histogram as color descriptor, GLCM (gray level co-occurrence matrix) as texture descriptor and edge histogram as edge descriptor. The training is carried out using backpropagation algorithm. This trained when presented with a query image retrieves and displays the images which are relevant and similar to query from the database. The results show a considerable improvement in terms of precision and recall of image retrieval. An average retrieval precision of about 88% and an average recall rate of about 78% is achieved using the proposed approach over SIMPLIcity project database.