摘要
针对目前单纯依赖于分析图像内容或文本关键词的成人图像判定算法的不足,提出一种融合网络图像的相关文本特征与图像内容语义特征的成人图像判定算法。成人图像的特征信息可能存在于其图像内容及其相关文本如图像文件名、所在网页中。在视觉词袋模型的基础上,将文本分析得到的相关文本特征与图像视觉元素特征如纹理、局部形态等进行底层特征融合,并采用支持向量机分类器实现图像分类。实验结果表明,该算法具有较好的分类效果。
In view of the disadvantage of the present adult image recognition algorithms such as ones only depend on the image content or keywords of text, this paper proposes a new adult web images recognition algorithm fusing the features of relative text and image semantic. Adult image feature information may exist in its image content and its associated text,such as the image file name or the webpage. Based on Bag-of-Visual-Words model, this algorithm coalesce text features extracted by analysis of corresponding text and visual features by analysis of image content such as texture, local shape on the low-level, and accomplishes image classification by using SVM classifier. The experimental results demonstrate effective classification capability of this algorithm.
出处
《计算机工程与应用》
CSCD
北大核心
2015年第4期175-179,208,共6页
Computer Engineering and Applications
基金
山东省自然科学基金计划项目(No..ZR2012FM021)
关键词
成人图像
图像识别
视觉词袋
文本分析
SVM分类
pornographic image
image recognition
Bag-of-Visual-Words
text analysis
SVM classification
