一种基于色阶偏差的皮肤检测技术研究

为提高数字图像中皮肤检测的准确性，在研究数字图像皮肤纹理特点的基础上，分析了基于灰度统计值的纹理检测算法及问题，在RGB彩色空间将基于灰度直方图统计矩的方法改造成基于纹理刷的色阶偏差法，并构建皮肤纹理检测器，对肤色检测器检测输出的皮肤区域再做纹理检测，保留和皮肤纹理一致的区域，去除误检区域。实验结果表明，该方案可提高对图像皮肤区域检测的有效性。

An Approach of Skin Detection Based on the Color-order Deviation

In order to improve the efficiency of detecting the skin areas in digital images, the characteristics of skin texture was studied and the method of skin texture detection based on the statistical moment of grey level histogram was proposed. The method based on the statistical moment of grey level histogram was then transfered into the color-order deviation one based on the texture-brush in RGB color space. A skin texture examining tool was designed by re-detecting the texture of certain skin areas detected by the skin color detection tool so as to remain the part accorded with the texture features and wipe off the incorrect. The result showed that the efficiency of detecting the skin areas in images is improved.