Texture classification using optimal Gabor filters
M. Pakdel, Farshad Tajeripour · 2011
Texture classification plays an important role in computer vision and its applications. Among various feature extraction method, filter bank method such as Gabor filters has emerged as one of the most popular one. This filter bank is defined by its parameters including frequencies, orientations, frequency ratio and smooth parameters of Gaussian envelope. In texture classification, Gabor filters show a strong dependence on a certain number of parameters, thus its performance depends on the selection of proper set of values for filter parameters. Also, the large number of filters leads to expensive computation in classification, thus it is necessary to perform a feature selection scheme to identify subset of effective and discriminate filters. In the present study, we first compute optimal Gabor filter parameters for texture classification, based on a kind of Genetic Algorithm and next a new method for filter selection is proposed. Classification accuracy and number of used filters on standard dataset, indicate efficiency of the proposed approach.