Supervised Classification of Texture Patterns with Nonnegative Matrix Factorization

Nonnegative Matrix Factorization (NMF) is an efficient tool for clustering and supervised classification of various objects, including text document, musical recordings, gene expressions, and images. In this paper, we are concerned with supervised classification of texture patterns. NMF is used for creating localized nonnegative feature vectors and low-dimensional nonnegative encoding vectors from a set of scale and rotation invariant descriptors of key-points that are extracted from the training images. We apply interior-point and active-set methods for estimating the nonnegative factors in NMF. The classification experiments for the selected images taken from the UIUC database demonstrate a high efficiency of the discussed approach.