陳劍軍

摘 要 基于矩陣乃“局部構成整體”的思想和并行計算的模式，將矩陣分塊進行非負矩陣分解，并將其用于圖像壓縮。實驗表明：該方法可減少存儲量、計算量，計算量的減少較為顯著。

關鍵詞 矩陣分塊 矩陣Hadamard乘積 NMF 圖像壓縮

中圖分類號：TN911.73 文獻標識碼：A DOI：10.16400/j.cnki.kjdkx.2016.09.066

Abstract Based on the idea of "local integral whole" and the parallel computing model， the matrix is divided into non negative matrix factorization， and it is used for image compression. Experimental results show that this method can reduce the amount of storage and computation， and the computation is more significant.

Key words Matrix block； matrix； Hadamard product； NMF； image compression

參考文獻

[1] Lee D.D，Seung H.S.Learning the parts of objects by non-negative matrix factorization[J].Nature， 1999， 401 （6755）：788-791.

[2] CHENG Ming-song，LIU Shao-lian.A Practical Fast NMF Algorithm[J].Journal of Dalian University of Technology，2013，53（1）： 151-156.

[3] GAO Hong-tao.Study on Theory and its Application of Non-negative Matrix Factorization Algorithm[D]. Shanghai： Tongji University.2005：8-18.

[4] WANG Xuan-sheng，CHEN Zheng，LU Lin-zhang. Lanczos Bidiagonalization： A Fast Start for Non-negative Matrix Factorization [J].Journal of Xiamen University（Natural Science），2012，51（2）：149-152.

[5] XU Sen，LU Zhi-mao，GU Guo-chang.Integrating K-Means and Non-negative Matrix Factorization to Ensemble Document Clustering [J].Journal of Jinlin University（Engineering and Technology Edition），2011，41（4）：149-152.

[6] LI Le，ZHANG Yu-jin.A Survey of Non-negative Matrix Factorization[J]. ACTA ELECTRONICA SINICA[J].2008，36（4）：738-743.

[7] XU Tai-yan，HAO Yu-long.The Current Research Situation Analysis of Non-negative Matrix Factorization and Applications[J].Journal of Wuhan Polytechnic University， 2010，29（1）：109-114.

[8] Wild S， Curry J，Dougherty A.Improving non-negative matrix factorization through structured initialization [J].Pattern Recognition， 2004，37（11）： 2217-2232.