Object based image ranking using neural networks

In this paper an object-based image ranking is performed using both supervised and unsupervised neural networks. The features are extracted based on the moment invariants, the run length, and a composite method. This paper also introduces a likeness parameter, namely a similarity measure using the weights of the neural networks. The experimental results show that the performance of image retrieval depends on the method of feature extraction, types of teaming, the values of the parameters of the neural networks, and the databases including query set. The best performance is achieved using supervised neural networks for internal query set.