基于多分类器组合的多角度彩色人脸图像检测

提出了基于神经网络和隐马尔可夫模型组合的彩色人脸图像检测方法 .根据归一化后的彩色图像的色度彩色分量直方图将图像粗分割成若干幅二值图像 ;在亮度图像上 ,以上述二值图像为掩模进行多分辨率的旋转不变性人脸检测 .在人脸检测时 ,本文分两步 :第一步先用神经网络来确定人脸的旋转角度 ,然后对旋正后的图像运用识别人脸奇异值特征的隐马尔可夫模型进行验证 .实验结果表明 ,本文算法是有效的

Rotation-Invariant Face Detection in Color Images Based on Multiple Classifiers

Presented a novel algorithm for rotation invariant face detection in color images, which combines both the capability of Neural networks and hidden Markov models (HMMs). Firstly, the color image is coarsely segmented into several binary images, which is accomplished based on the analysis of the histogram of the H component of the HSI color model. Secondly, multi resolution rotation invariant face detection is conducted in the luminance component image, while the previously segmented binary images are employed as masks. And lastly, both Neural networks and HMMs are utilized sequentially to decide whether a sub image is a face, i.e., the Neural networks tells about the rotation angle while the HMMs verify the de rotated upright face. Both the chromatic and gray information of a face are fully explored in our system. The performance of the system has been tested over 200 test images of varying complexity, including scanned photos, Internet images, and cluttered scenes captured in movies, with promising results, which have proved that the proposed algorithm is effective.