基于非负矩阵分解的图像特征提取技术-中国科技论文在线.PDFVIP

** 150001 5 (non-negative matrix factorization, NMF) , , , 10 Non-negative Matrix Factorization Algorithms for Image 15 Extraction Wang Kejun, Zuo Chunting (Harbin Engineering University,Harbin 150001) Abstract: Non-negative matrix factorization (NMF) is a new factor extraction method for pattern recognition. It aims to find two non-negative matrices whose product can well approximate the 20 original matrix, which naturally leads to parts-based representation. This paper firstly describes the basic idea of non-negative matrix factorization, provides the objective functions and the iterative algorithm, and then take finger vein images as an example of its application to extract features. In next part, it puts even more weight to discuss the available improved approaches and draws some conclusion on the future improvement. 25 Key words: pattern recognition; non-negative matrix factorization; feature extraction 0 [1] [2,3] [4] 30 Principal Component Analysis PCA Independent Component Analysis [6-8] ICA 35 Lee Seung 1999 —— [4] non-negative matrix factorization NMF NMF NMF 40 () [