Illumination invariant recognition of color texture using correlation and covariance functions

In this paper, we derive a complete set of Zernike moment correlation functions 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. Circular and non-circular correlations are discussed and comparisons with a previously suggested color covariance functions have been carried out using about 600 different illuminations and rotations textured images. Using moment correlation matrices in the invariant recognition of color texture, the process can promise in high computation efficiency as well as recognition accuracy. The derived correlation invariants is proposed as a general formalism that can be used directly with other kinds of complex moments, e.g. Fourier Mellin, pseudo Zemike, disc-harmonic coefficients, and wavelet moments, to obtain moment correlation based invariants.