This paper discusses a method of improving the discrimination power of a certain class of GLCM features. We investigate where co-occurrence matrix features derive their discriminatory power, and provide a theoretical basis for improving this method. Finally, we present examples of discrimination improvement using real-world data. Cross-validation results have indicated remarkable increases in feature discriminatory power for almost all features trialed. The co-occurrence feature Variance was a good example, with an average 70% decrease in misclassification after implementing the improvements detailed in this paper