Abstract: This paper presents a skin region segmentation algorithm for face recognition in conditions of varying illumination, multiple pose and complex background. This algorithm preprocesses the color facial input image to correct the color cast and adjust the brightness, then it builds an adaptive sphere skin model, based on which it calculates the adaptive skin similarity, finally it segments and extracts the skin region object using the skin similarity and adaptive dynamic threshold. Experimental results show that the algorithm has good segmentation precision and adaptability.

Key words: pattern recognition, skin color segmentation, similarity, skin model

摘要： 提出一种用于变化光照、多姿态和复杂背景条件下人脸识别的肤色区域动态分割算法。对彩色人脸输入图进行色偏校正和亮度调节预处理，利用肤色聚类特性构建一种自适应球体肤色模型，并基于该模型计算自适应肤色相似度，利用肤色相似度，采用自适应的动态阈值进行肤色区域目标的分割和提取。实验结果表明，对于变化光照、多姿态和复杂背景的彩色人脸图像，该算法有良好的分割精度和自适应性。

关键词: 模式识别, 肤色分割, 相似度, 肤色模型