Designing Multiple Gabor Filters for Multi-Texture Image Segmentation

We consider the problem of segmenting multitextured images using multiple Gabor filters. In particular, we present a mathematical framework for a multichannel texture-segmentation system consisting of a parallel bank of filter channels, a vector classifier stage, and a postprocessing stage. The framework establishes mathematical relationships between the predicted texture-segmentation error, the frequency spectra of constituent textures, and the parameters of the filter channels. The framework also permits the systematic formulation of filter-design procedures and provides predicted vector output statistics that are useful for classifier design. This paper focuses on the mathematical framework and provides experimental results that confirm the utility of the framework in the design of a complete image-segmentation system. The results demonstrate effective segmentation using a straightforward classifier and fewer than half the number of filters needed in previously proposed approaches. Subject terms: Gabor prefilter, Gabor filter, Gabor function, texture segmentation, statistical image analysis, texture analysis, computer vision, image segmentation.