【作者】 李国军；

【导师】 沈永增；

【作者基本信息】 浙江工业大学 ， 控制理论与控制工程， 2003， 硕士

【摘要】 人脸自动识别技术有广阔的应用前景和迫切的现实要求，是当前模式识别领域最热门的研究方向之一。人脸是一种复杂的模式，从背景中定位人脸是一件非常困难的事，其难度不小于人脸识别本身。研究人员对人脸识别作了大量的研究，但对复杂图像人脸定位的研究却不多。本文首先对人脸定位和人脸识别的意义、人脸定位和人脸识别的发展作了概括性的介绍；然后对一些基本理论作了简单的介绍，紧接着对人脸定位的方法做了全面的论述。 本文第四章提出了一种基于简单背景的人眼定位算法，用小波变换对人脸图像进行边缘检测，对得到的结果进行行列统计，得到人脸的大致位置，再用模板匹配进行最后的精确定位。 在本文最后提出了一种在静态图像中定位出人脸的方法，此方法对图像没有任何限制，对人脸的个数、大小、位置、方向也没有任何限制。具体就是利用K-L变换将肤色样本和非肤色样本的RGB三种颜色分量重新构造了一个颜色空间，肤色空间和非肤色空间在该空间的距离得到拉伸，因此，两类样本具有良好的可分性，从而实现了肤色区域和非肤色区域的分割。最后，利用自定义的模板来实现人脸的定位。通过实验仿真证明，此方法能够较好的实现了人脸定位。更多还原

【Abstract】 Automatic human face recognition is attractive in pattern recognition and image procession. The human face is a complex pattern. It is very diffcult that locate face in complex background. It is even harder than face recognition. Many researches have been done upon face recognition, more effort need to be directed in face location problem. In this paper , first , a survey of human face location and recognition is given. Then, it introduces the developments of the human location and recognition. Afterword, it discussed the basis theory and method.In chapter IV , a new method of eyes location is provided. It detects the edg of human face image with wavelet transform in the first. Then it begins to statistic the gray value of the row and list. So we can get approximate position. At last we can get accurate position with template maching.At last, a method of location face in the static image is provided without any limitation upon images , or any condition about number, size, position and orientation of faces. A new color space is drived from RGB color format of the skin colors and other background colors with K~L transforms. The distance between the skin color and the background colors is increased in this color space. So we can segment the skin regions from one image. Then we match with templates that was difined by myself. At last we can get the position of the human face. The experiment result shows that this method finishes the human faces location.更多还原