Designing Gabor filters for optimal texture separability

The discrimination of textures is a critical aspect of identification in digital imagery. Texture features generated by Gabor filters have been increasingly considered and applied to image analysis. Here, a comprehensive classification and segmentation comparison of different techniques used to produce texture features using Gabor filters is presented. These techniques are based on existing implementations as well as new, innovative methods. The functional characterization of the filters as well as feature extraction based on the raw filter outputs are both considered. Overall, using the Gabor filter magnitude response given a frequency bandwidth and spacing of one octave and orientation bandwidth and spacing of 303 augmented by a measure of the texture complexity generated preferred results.