基于双正交小波的彩色人脸检测算法的研究

该算法将分水岭算法与小波变换相结合，首先利用分水岭算法对彩色图像进行基于肤色特征的预处理，确定类人脸区域，再用双正交小波对于处理后的图像进行边缘检测，并用椭圆法验证得到待定人类区域是否为人脸区域，将两次结果相比，最终确定人脸区域，本算法充分利用了分水岭算法图形分割边缘连续、速度快、精度高的特点，同时结合双正交小波基正交条件弱的优势，弥补了分水岭算法过分割的缺点，从而既能够快速有效地检测出人脸又能够减小图像的存储空间。

Research on Color Face Detection Based on Biorthogonal Wavelet

This arithmetic uses Watershed arithmetic combined with wavelet. Firstly, the watershed preprocess image based on skin color character and conform some areas as maybe face area, then uses biorthogonal wavelet do edge detection to ensure the maybe face areas whether a face or not. The arithmetic embodies the merits of watershed just like image edge constant, high speed, high precision and so on. At the same time, uses Biorthogonal Wavelet to fetch up disadvantage about the over--segmentation of standard watersbed segmen ration, therefor, a face on an image can be detected fast and efficiently.