摘要
利用肤色信息在彩色图像中检测人脸速度快、易于实现,但准确率不高。本文提出首先应用人脸肤色信息在彩色图像中进行人脸粗略检测,得到候选人脸区域,再以基于梯度模板匹配的方法进行人脸精确定位,从候选区中准确地找到人脸。将这两种方法结合在保证精确的人脸检测的前提下,可有效提高检测速度。
Using skin color information to detect human face in colorful image is quick and practicable, but is not accurate. The paper proposes a new method that firstly skin color is applied to detect face roughly and get candidate regions of face, and then the algorithm based on gradient template matching is used to locate human face accurately from candidate regions. The proposed method can not only detect human face from complicated background, but also effectively increase detecting speed.
出处
《自动化与仪器仪表》
2008年第6期91-93,105,共4页
Automation & Instrumentation
基金
高等学校博士学位学科点专项科研基金
关键词
人脸检测
肤色信息
模板匹配
Human face detection
Skin color information grads
Template matching
