一种与局部模型相结合的自适应肤色检测方法

使用预先训练得到的肤色高斯模型进行肤色检测，容易受到环境影响。为了提高肤色检测的性能和精度，提出一种与局部模型相结合的自适应肤色模型检测方法。在YCgCr颜色空间中使用高斯模型对图像进行肤色分割，并利用初次检测到的肤色区域像素数据建立局部模型，估计出模型参数。把局部模型与预先训练得到的参数相结合，得到自适应的肤色模型。实验表明，该方法和单独的肤色高斯模型相比，提高了在不同背景和光照下肤色检测正确率。

Adaptive skin detection method combining local skin color model

Using pre-trained?Gaussian model?for?skin?color?detection is easily affected by?environment.?In order to improve?the performance and accuracy of?skin detection, an adaptive skin?model combining?with the?local skin color?model?method is proposed.?Based on YCgCr?color space, the Gaussian model?is used to?detect the skin color region, and the?detected?color?pixel?data is used to establish?a local?model and estimate the model’s parameters. Combining the pre-trained model’s parameters and local model’s parameters, this article obtains an adaptive skin model.?Experiments?show that the method and the?single skin color Gaussian model?compared to?improve probability of correct detection in the different light environment?and background.