Artificial Neural Networks as Decision-Makers for Stereo Matching

Abstract

This paper investigates the use of artificial neural networks to help making a decision on matching of stereo images. An image matching technique based on extracting features from segmented regions is adopted in this work, and a neural network framework is applied for region matching of stereo photographs. Two types of neural networks are used, the radial basis network, (RB) for learning clustering, and the back propagation (BP) network for learning image matching. The (RB) neural network is to cluster the regions according to the locations of their centered points. For each region, the BP network uses differential features as input training data. While training and testing the system, multiple features are extracted and used for enhancing the accuracy of the matching process. Features include (compactness, Euler number, and invariant moments) for each region. Results obtained from the neural networks (namely; clustering and initial matching array) are used to select the best matching pair. Results are showing a good matching accuracy.