A New CBIR System Using SIFT Combined with Neural Network and Graph-Based Segmentation

In this paper, we introduce a new content-based image retrieval (CBIR) system using SIFT combined with neural network and Graph-based segmentation technique. Like most CBIR systems, our system performs three main tasks: extracting image features, training data and retrieving images. In the task of image features extracting, we used our new mean SIFT features after segmenting image into objects using a graph-based method. We trained our data using neural network technique. Before the training step, we clustered our data using both supervised and unsupervised methods. Finally, we used individual object-based and multi object-based methods to retrieve images. In the experiments, we have tested our system to a database of 4848 images of 5 different categories with 400 other images as test queries. In addition, we compared our system to LIRE demo application using the same test set.