A complete set of Zernike moment correlation functions is used to capture spatial structure of a color texture. The set of moment correlation functions is grouped into moment correlation matrices to be used in illumination invariant recognition of color texture. For any change in the illumination, the moment correlation matrices are related by a linear transformation. Comparisons between suggested color covariance functions with circular and non circular correlations have been carried out using about 600 textured images in different illuminations and rotations conditions. Our suggested method can promise in high computation efficiency as well as recognition accuracy.