基于霍夫曼树SVM的监控视频清晰度评价  

Sharpness Recognition of Surveillance Video based on Huffman-tree SVM

作　　者：翟旭[1] 戚玲[2] 喻松[1] 

机构地区：[1]北京邮电大学信息光子学与光通信研究院,北京100876 [2]北京邮电大学信息与通信工程学院,北京100876

出　　处：《软件》2012年第12期126-129,共4页Software

摘　　要：为提高视频监控系统中视频清晰度评价的准确性和系统运行效率,提出了结合提升小波变换和支持向量机(SVM)分类算法的图像清晰度评价算法。在研究了常用的SVM多类分类算法的基础上,提出了霍夫曼树SVM多类分类算法。进行了视频监控系统的仿真实验,结果表明,与普通二叉树SVM多类分类算法的效果相比,霍夫曼树SVM多类分类算法在系统的分类正确率及运行效率方面有明显提升。To increase the accuracy and operational efficiency in evaluating the sharpness of digital video of the video surveillance system, the digital image quality recognition algorithm, which combined lifting wavelet and Support Vector Machine （SVM）, was proposed. After the common SVM multi-class classification algorithms were analyzed, the Huffman-tree SVM multi-class classification algorithm was proposed. Experiments simulating the Video Surveillance System were taken, and the results, comparing to the binary-tree SVM multi-class classification algorithm, indicated that the new method speeded up the running and yielded higher precision.

关 键 词：图像处理 清晰度评价 提升小波变换 支持向量机 霍夫曼树 

分 类 号：TN919.82[电子电信—通信与信息系统]