基于YCbCr颜色空间的不良图片检测研究  

Indecent Images Detection Based on YCbCr Color Space Research

作　　者：胡柳[1] 周立前[1] 

机构地区：[1]湖南工业大学计算机与通信学院,株洲412008

出　　处：《科学技术与工程》2013年第15期4433-4436,共4页Science Technology and Engineering

摘　　要：为了实时监测网页中的不良图片,提出一种以YCbCr颜色空间肤色检测为主,结合模板匹配、特定情景的识别实现对不良图片的快速检测,采取提高肤色检测正确率、消除背景区域及模板匹配等过程逐步明确其类别。通过选取网络图片进行实验,其中正常图片的识别率为90.3%,不良图片识别率为89.5%,表明该方法能应用于实际过滤系统中。In order to indecent images of real-time monitoring of Web pages, YCbCr color space is proposed mainly for skin detection, combined with the template matching, identification of specific scenarios to achieve the rapid detection of bad pictures taken to improve the correct rate of skin detection, and eliminate background region and the template matching process gradually clear category. Through selected network picture experiment, which the normal picture identification rate was 90. 3 %, the bad picture recognition rate of 89.5 %, show that the method can be applied to the actual filtration system.

关 键 词：不良图片 YCBCR颜色空间 模板匹配 领域信息 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP302[自动化与计算机技术—计算机科学与技术]