Using Gabor filter for the illumination invariant recognition of color texture

We derive Gabor feature matrices that are used to recognize color texture independent of illumination changes. The feature depends on Gabor energy feature in which we derive a multi-band Gabor energy feature matrix. We also prove that two Gabor energy feature matrices such that represent two corresponding textures under different illumination are related by a linear transform. Experimental results show that the model is successful in recognizing color texture independent of the change in the spectral illuminations.