Using Neural Networks and Self-Organizing Maps for Image Connecting

This paper addresses the problem of connecting a sequence of images acquired by a camera rotating about its center. A novel method is introduced using neural networks. The algorithm finds the major overlapping area in the images to be connected using the neural network and then joins the images. The inputs to the network are edge detected images created by applying an algorithm on the original images. The paper presents a theoretical and computational investigation into connecting any two given images using a Self-Organizing Map (SOM). Simulation results demonstrate that self-organizing neural networks can be efficiently used for this purpose.