复杂背景彩色图像中多角度人脸检测

Rotation-invariant Face Detection in Color Images with Complex Background

提出一种针对复杂背景彩色图像中的人脸检测方法。基于GrayWorld假设,在RGB颜色空间采用颜色平衡的方法对偏色图像进行颜色校正,在YES颜色空间进行肤色检测并应用预处理技术缩小人脸检测的搜索区域。在物体区域方向计算的基础上,提出能够检测任意旋转角度人脸的方法,在人脸候选区域采用模板脸匹配方法定位人脸。实验表明,该方法对不同光照环境、不同尺寸、任意旋转的人脸有较好的检测效果。

This paper presents a method for face detection in color images with complex backgrounds. Which adopts color balance method to correct the images color bias in the RGB color space based on ＂Gray World＂ hypothesis; detects skin color in the YES color space, introduced preprocessing techniques to reduce the searching areas of the face detection, and proposes a method that can detect arbitrary rotated faces based on the computation of the direction of the object. Template-matching method is used in the face candidate regions to locate face. The experiments show that the proposed face detection method provides better detection results to different illumination, different sizes and arbitrary rotated faces.