A Comparison of Colour Texture Attributes Selected

In this paper two methods for selecting input features for a neural network used to aid iconic retrieval in an image database are presented and compared. The rst method involves training the network on all the feature inputs and then analysing the weight values in an attempt to nd the more important input features. The second borrows a method from statistical feature selection known as the sequential oating forward selection algorithm. Both methods improve the computationally eeciency of the image database search whilst not aaecting the quality of the results obtained. Being more computationally eecient, network weight analysis ooers an attractive alternative to the statistical method.