Reducing the Dimension of Color Features Using a Naive Bayesian Classifier

Color histograms are usually used as the color feature vectors for classifying the color of objects in images. We reduce the dimension of the feature vector by a factor of about 30 by using a naive Bayesian classifier, and use the resulting feature vectors with a support vector machine to recognize vehicle colors. Experiments show that the recognition rate is close to that achieved with the original large feature vectors, while recognition time is reduced by a factor of more than 30. We also show that our method outperforms principal component analysis.