Zhang, Jiang; Wang, Nian; Liang, Dong; Tang, Jun; Zhou, Meiju Image classification based on non-negative matrix factorization and adjacency spectra. (Chinese. English summary) Zbl 1174.68575 J. Univ. Sci. Technol. China 38, No. 3, 247-251 (2008). Summary: Combined Non-negative Matrix Factorization (NMF) with adjacency spectra, a new method of image classification is proposed to extract characteristic information of an image. Firstly, the adjacency matrix is constructed by the feature points of the image. Secondly, the initial value of NMF iteratives is evaluated by means of adjacency matrix, and then the samples of image classification are obtained through basis vectors of NMF. Finally, image classification is performed by adopting probabilistic neural networks classifier. Experimental results of synthetic data and real images show that the method not only has feasibility and validity, but also further improves recognition rate and stability of image classification. MSC: 68T10 Pattern recognition, speech recognition 68U10 Computing methodologies for image processing Keywords:non-negative matrix factorization; adjacency spectra; graph spectrum; image classification PDFBibTeX XMLCite \textit{J. Zhang} et al., J. Univ. Sci. Technol. China 38, No. 3, 247--251 (2008; Zbl 1174.68575)