Feature Fusion with Neighborhood-Oscillating Tabu Search for Oriented Texture Classification

This paper develops two techniques of oriented texture analysis: the modified Gabor filters (MGF) and the Gaussian Markov random field model with circular neighborhoods (CGMRF). The neighborhood-oscillating tabu search algorithm (NOTS) is proposed to solve the MGF/CGMRF feature fusion problem, and compared with classical algorithms, such as sequential forward selection and sequential forward floating selection methods. Based on the experimental results, NOTS is shown to be a promising tool for feature fusion, and the MGF/CGMRF fused features achieved by NOTS perform better than either MGF or CGMRF alone according to the Fisher criterion and classification accuracy.