人脸检测中的肤色提取模型研究

肤色是人脸区别于其他景物的一个重要特征，基于肤色的人脸检测第一个要解决的问题就是选择合适的肤色提取模型，模型选择的好坏直接影响肤色提取效果，从而影响人脸检测的准确性。该文就近期人脸检测中常用的肤色提取模型进行了简单地分析与比较表明，色度空间中的亮度分量是造成各种模型肤色提取效果不同的主要原因，并针对Anil K．Jain的Cb、Cr椭圆聚类法肤色分割的缺点进行了改进，首先对亮度信息设置阈值，进行分段判断，这样克服了在高亮度区域和亮度较低的区域中存在的不足。实验证明，该文的方法较之其他方法效果要好，在人的眼睛、嘴巴等部分有更好更细致的检测效果。

Skin Pick-up Patterns for Face Detection

The skin color is an important characteristic different from other sceneries, the first problem of the face segmentation based on skin color is to choose the appropriate skin pick - up patterns, and the patterns determine skin pick - up result,thus affecting the accuracy of face detection. In this paper, a simple analysis and comparison according to the skin pick - up patterns of face detection shows that the luminance information is the primary factor that makes different results of various skin pick - up patterns, and the Cb Cr ellipse clustering method of Anil K. Jain is improved by establishing a threshold value of luminance information and carrying out a divided judgment. The experiments show that the improved method is better than other methods and much better in distilling the person' s facial features.