摘要： 对于有色偏、高光和阴影的复杂背景彩色人脸图像，已有模型不能获得较理想的分割结果。为此，提出一种基于CbCgCr椭球体肤色似然平滑度的肤色分割算法。预处理人脸输入图，以消除色偏、高光和阴影，建立CbCgCr椭球体肤色模型，计算肤色似然平滑度，利用肤色似然平滑度粗分割肤色区域，并对其进行去噪处理，以获得肤色细分割图和细提取图。实验结果表明，对光照不均和背景复杂的人脸彩色图像，该算法的肤色分割准确性、鲁棒性和实时性较优。

关键词: 生物特征, 模式识别, 肤色分割, 似然, 平滑度, 椭球体模型

Abstract: For complex background color face image with color bias, high light and shadow, existing model cannot get satisfactory segmentation results. In order to solve this problem, this paper presents a skin color segmentation algorithm based on CbCgCr ellipsoid likelihood smoothness. This paper segments the skin region using smoothness of CbCgCr ellipsoid skin likelihood. It preprocesses the facial input image for eliminating the color bias, high light and shadow, calculates the smoothness of skin color likelihood of facial input image after processing by building on CbCgCr ellipsoid skin color model, and then roughly segments the skin region by using the smoothness of skin color likelihood, removes the noise of skin color coarse segmentation image and gets the skin color fine segmented image and fine extracted image. Experimental results show that the algorithm has good results and robustness of skin color segmentation for color facial image with unequal illumination and complex background.

Key words: biological feature, pattern recognition, skin color segmentation, likelihood, smoothness, ellipsoid model

中图分类号: 

• TP301.6