基于自适应光线补偿的人脸检测算法

Face Detection Algorithm Based on Self-adaptive Light Compensation

人脸检测是计算机视觉、模式识别、人机交互中一个很受关注的研究热点。一般的肤色检测算法假定待测图像是在均匀光照下获得的,环境光照变化、光照不均必会影响图像的柔和度,进而影响人脸的正确检测。针对这一问题,提出一种自适应光线补偿的人脸检测算法,应用不同光照情况下拍摄的69幅彩色图像进行对比实验研究。实验结果表明:该方法有效可行,检测率达96.64%,比单纯应用肤色模型法提高检测率7个百分点,降低误检率5个百分点。

Face detection is a hotspot on the research of computer vision, pattern recognition and human-computer mutuality. General face detection algorithms based on skin color assumes that the color images were obtained on even illumination condition, so the uneven illumination distribution must affect the sofmess of the color images, which will result in detection error. In order to solve this problem, a new face detection algorithm based on self- adaptive light compensation is proposed in this paper. With 69 color pictures taken by different illumination as detection data, the method is tested. The result shows that this method, which can improve the detection rate 7 % （detection accuracy is 96.64 % ）and reduce the false detection rate 5 % than general face detection algorithms, is effectual and feasible.