Efficient Gabor filter design for texture segmentation

Gabor lters have been successfully applied to a broad range of image processing tasks. The present paper considers the design of a single lter to segment a two-texture image. A new e cient algorithm for Gaborlter design is presented, along with methods for estimating lter output statistics. The algorithm draws upon previous results that showed that the output of a Gaborltered texture is modeled well by a Rician distribution. A measure of the total output power is used to select the center frequency of the lter and is used to estimate the Rician statistics of the Gaborltered image. The method is further generalized to include the statistics of post ltered outputs that are generated by a Gaussian ltering operation following the Gabor lter. The new method typically requires an order of magnitude less computation to design a lter than a previously proposed method. Experimental results demonstrate the e cacy of the method. Image segmentation Texture Gabor lters Rician statistics Texture segmentation Image statistics Texture analysis