Color-segmentation is very sensitive to changes in the intensity of light. Many algorithms do not tolerate variations in color hue which correspond, in fact, to the same object. Learning Vector Quantization (LVQ) networks learn to recognize groups of similar input vectors in such a way that neurons physically near to each other in the neuron layer respond to similar input vectors. Learning is supervised, the inputs vectors into target classes are chosen by the user. In this work a new algorithm based on LVQ is presented. It involves neural networks that operate directly on the image pixels with a decision function. This algorithm has been applied to spotting and tracking human faces, and shows more robustness than other algorithms for the same task.
