Trainable Look-Up-Tables versus Neural Networks for Real-time Colour Classification

Abstract

The pixel-wise classification of CCD colour images into previously learned colour classes at video-rate is a demanding vision task, both with regard to the complicated cluster shapes encountered in natural scenes and to the required computing power for real-time operation. We discuss the use of a perceptron Neural Network and propose an alternative simple and low-cost classifier based on approbriately trained look-up-tables. Two different learning rules for the supervised training of this LUT classifier are presented. This LUT classifier shows all the positive features of a 3-layer perceptron Neural Network, but performs 60.000 times faster then the Neural Network PC-simulation.