基于纹理的皮肤检测

Texture-based Skin Detection

实时敏感图像检测可以有效地控制国际互连网中的色情图像传播。它的检测手段主要是对皮肤区域进行辨识,而皮肤纹理检测又是皮肤区域辨识中的重要内容。该文采用基于DCT变换和Gabor小波变换两种方法进行皮肤纹理的特征提取,提取的特征作为高斯混合模型(GMM)输入向量,然后通过GMM进行皮肤纹理检测。实验是在360幅敏感图像以及2400幅经过分类的非敏感图像的测试集进行,结果表明基于Gabor小波变换的皮肤纹理检测取得了比基于DCT变换更好的效果,其正检率可以达到95%。

Real-time detection of pornographic images can effectively prevent pornographic images from spreading on the Internet.Its detection methods mainly focus on the identification of skin region in which skin texture detection plays an important role.In this paper,two methods based on DCT and Gabor wavelet transform are respectively proposed to extract the features of skin texture.The extracted features are inputted into Gaussian mixture model(GMM)with which the skin texture detection is performed.Experiments are carried on a test set of360pornographic images and2400as-sorted control images.Experimental results show that Gabor wavelet transform based skin texture detection has better performance than DCT-based,and its correct detection rate can reach95%.