Multicollinearity exists in texture measures, implying that the dispersion matrices are nearly singular. The generalized inverse was employed to solve this problem, yielding an 8% improvement in classification accuracy. © 1979.
Hsu, S. Y. (1979). The Mahalanobis classifier with the generalized inverse approach for automated analysis of imagery texture data. Computer Graphics and Image Processing, 9(2), 117–134. https://doi.org/10.1016/0146-664X(79)90052-2