The classification of table olive in different quality categories is performed depending on the defects in the surface of the fruits. However, the characteristics of every category are not defined. Then, it is necessary to apply learning algorithms that allow the extraction of quality information from batches previously classified by expert workers. In this research, a colorimetric characterisation of the more common defects has been carried out. An image analysis system has been used to segment the parameter set with the information from the olives quality. Three different algorithms have been applied to classify the olives in four quality categories. The results show that a neural network with a hidden layer is able to classify the olives with an accuracy of over 90%, while partial least squares discriminant and Mahalanobis distance are over 70%. (C) 2003 Elsevier Ltd. All rights reserved.