基于显著区域的敏感图像识别方法  

Recognition algorithm of sensitive image based on salient region

作　　者：王国营[1] 康进峰[1] 

机构地区：[1]解放军信息工程大学电子技术学院,河南郑州450004

出　　处：《计算机工程与设计》2011年第1期236-239,共4页Computer Engineering and Design

摘　　要：依靠先进的技术识别和过滤网络上大量的色情信息正是时下政府相关管理部门急需解决的问题,而敏感(色情)图像在网络色情信息中占有相当大的比例。依据敏感图像的显著区域具有丰富的肤色信息这一特点,利用一种基于肤色信息的投票机制提取出显著区域,在显著区域上进行特征提取,最后采用支持向量机的方法对图像进行分类,从而完成敏感图像的识别。实验结果表明,该方法的正检率达到88.2%,能有效地识别出敏感图像。To solve the urgent problem that identifying and filtering rich erotic information in the internet depend on advanced technology for the supervision department of government,a recognition algorithm of sensitive image based on salient region is proposed.Salient regions of erotic images have rich skin information.A voting mechanism based on the skin information is used to extract salient region. Then,feature extraction is carried out in the salient region.Finally,support vector machines is used to classify images.The experimental results show that the sensitive image recognition algorithm provided reach at 88.2% precision rate and it is able to identify sensitive images effectively.

关 键 词：显著区域 肤色信息 敏感图像 特征提取 支持向量机 识别算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]