Performance comparison of feature extraction methods for neural network based object recognition
Claus Neubauer, Ming Fang · 2003
Pre-processing and feature extraction can significantly enhance the performance of a neural network based classifier. In this paper several feature extraction techniques including edge filters, local features and distance transformation are selected for image pre-processing in order to improve the recognition accuracy in combination with a neural network classifier. The visual object recognition performance of these algorithms is extensively compared based on a real world data set with significant variation of viewpoint and illumination.