Texture classification using gabor energy features and higher order spectral features: a comparative study
Ronald Tabu Elunai, Vinod Chandran, Srinitha Sridharan · 2006
Several approaches to the classification and segmentation of textural content in digital images have been investigated in recent years. The extraction of features for classification has particularly received considerable attention. In this paper we contrast between two approaches for feature extraction i.e. Gabor filters and bispectral invariant features. A subset of the Brodatz album are used in a dichotomy experiment and separate SVMs are trained to classify features from each pair. Our experiments show that the bispectral invariant features produce better classification results for more texture pairs than the Gabor filters. Results also indicate that a combination of the two feature sets will yield higher accuracy, and for some texture pairs, neither works very well.