Texture classification in different illumination conditions via testing the covariance matrices and mean vectors

Texture classification is of utmost importance in the image processing. In this paper the problem of texture classification is considered based on testing the covariance matrices and mean vectors. This allows us to determine the class of different images without the necessity of the training data. The generalized likelihood ratio (GLR) test is derived in order to classify several images. To make the classification robust to illuminance changes, we assume that the means of different images in one group, could differ by a constant value. Consequently the proposed test is invariant to the constant difference in the means of observations in each group. Computer simulations also confirm the efficiency of the classifier in dealing with the images with different illumination conditions.