基于图像内容及底层特征的敏感图像检测方法

内容安全是信息安全领域中的一个新兴且重要的研究方向.敏感图像分析与监控是该方向的一个重要研究分支,近年来愈来愈受到关注.然而,由于其研究起步较晚且图像本身所具有的复杂多样性,相对来说,目前已有的研究成果还比较少且性能上也有待于进一步提升,为此,本文提出了一种新的敏感图像分析检测方法.该方法从基于图像内容分析和基于图像底层特征分析两个方面进行综合分析检测,在基于内容分析方面,通过分析图像是否含有大量的皮肤区域来初步检测敏感图像;在基于图像底层特征分析方面,通过颜色直方图矢量和颜色聚合矢量提取图像的颜色及空间信息,并在此基础上利用SVM进行分类,从而实现敏感图像的初步检测;最后,对上述两方面的初步检测结果进行综合判决,实现敏感图像的最终检测.仿真实验结果出示了该新方法的有效性.

The Content and Bottom-Layer Feature Based Detection Approach for Nude Images

The content security is a new and important issue in information security. Erotic image analysis and monitoring is an important research branch of content security,and has been attracted more and more attention in recent years. However,due to the corre- sponding study only began several years ago and the images are complex,there are not many corresponding reported research results rela- tively and their performance needs improving further. Therefore,this paper proposed a new approach to analyze and detect the erotic ima- ges based on both image content analysis and image bottom-layer feature analysis. For the image content analysis,the detection of erotic image is performed by analyzing whether it contains large mass of skin region. For the image bottom-layer feature analysis,the detection of erotic image is performed by extracting the color and spatial information of the image based on color histogram vector (CHV) and col- or coherence vector (CCV) and making classification based on Support Vector Machines (SVM). The final detection result of the pro- posed approach is obtained by judging the above detection results integratedly. The experimental results show the effectiveness of the pro- posed approach.