NMF (non-negative matrix factorization) video compression technology



The invention discloses an NMF (non-negative matrix factorization) video compression technology which includes the steps of firstly, burning an embedded video compression with algorithm of recognition and separation technology, and loading the video compression into a chip; secondly, pretreating collected data by connecting the chip with equipment; thirdly, finding significant nodes through an NMF method in mode recognition; fourthly, realizing node saving and time domain cutting of real-time video streams through the significant nodes; and fifthly, separating background and foreground through similarity comparison, saving the background as a static image, and saving the foreground as sparse matrix data. By MATLAB (matrix laboratory) programming, the NMF theory and software-hardware combination, compression of time domain and space domain redundant information is achieved, algorithm is optimized, data storage is compressed greatly, and data processing speed is improved.




Download Full PDF Version (Non-Commercial Use)

Patent Citations (2)

    Publication numberPublication dateAssigneeTitle
    CN-101631237-AJanuary 20, 2010青岛海信网络科技股份有限公司一种视频监控数据存储管理系统
    US-2009271433-A1October 29, 2009Xerox CorporationClustering using non-negative matrix factorization on sparse graphs

NO-Patent Citations (1)

    李岳楠: "视频信息内容管理关键问题研究", 《中国博士学位论文全文数据库 信息科技辑》, 15 April 2011 (2011-04-15), pages 4 - 3

Cited By (2)

    Publication numberPublication dateAssigneeTitle
    CN-103793710-BFebruary 15, 2017中国科学院深圳先进技术研究院一种图像显著性物体检测方法和装置
    CN-103873879-AJune 18, 2014中山大学深圳研究院一种基于双重奇异值分解的视频图像压缩方法