基于分裂式K均值聚类的肤色检测方法

针对复杂图像背景及光照导致的肤色检测率不高的问题,提出一种基于分裂式K均值聚类的椭圆模型肤色检测方法。该方法对图像进行光线补偿处理,采用Gray World方法对图像进行颜色均衡,选择建立检测效率较高的椭圆肤色模型进行肤色检测,并在检测出的肤色区域上采用分裂式K均值聚类(FKM)进行二次的肤色判决,进一步准确检测出肤色区域。实验表明,所提出的检测算法能准确高效地检测出肤色区域,具有较高地准确率和较强的鲁棒性。

Skin color detection by fissive K-means clustering method

The complex image background and the varying lighting conditions lead to the low detection rate, this paper presents a novel color detection method to solve this problem. The Gray World color balance is carried out to compensating the light before detecting the skin area. After that choose to create the high detection efficiency ellipse model for skin color detection, then the paper detects skin color area on the K-means clustering using split（FKM）for the second sentence to make the detection more accurate. The experiment results show that this detection algorithm can detect skin area accurately and efficiently, and it has higher accuracy and robustness.