Comparing the Performance of Connectionist and Statistical Classifiers on an Image Segmentation Problem

In the development of an image segmentation system for real time image processing applications, we apply the classical decision analysis paradigm by viewing image segmentation as a pixel classifica.tion task. We use supervised training to derive a classifier for our system from a set of examples of a particular pixel classification problem. In this study, we test the suitability of a connectionist method against two statistical methods, Gaussian maximum likelihood classifier and first, second, and third degree polynomial classifiers, for the solution of a "real world" image segmentation problem taken from combustion research. Classifiers are derived using all three methods, and the performance of all of the classifiers on the training data set as well as on 3 separate entire test images is measured.