摘要
针对彩色图像中人脸在多角度状态下眼睛定位困难的现状,提出了一种基于颜色信息且不受头部姿势影响的人眼精确定位的新方法。该算法首先检测出彩色图像中的人脸位置;然后在人脸区域内,根据肤色与眼睛、眉毛等颜色的差异性进行区分,得到脸部区域的非皮肤部分;接着对非皮肤部分进行数学形态运算,去除干扰点;然后采用平均复杂度方法定位出眼睛区域,并采用相似度计算检验眼睛定位正确与否;最后在眼睛区域内,精确定位眼白、虹膜,通过计算眼睛面积分析眼睛睁开程度。实验结果表明,该算法能精确检测出彩色图像中的眼睛并能快速判断出眼睛的状态。该方法不受头部姿势影响,计算简单,运算速度快,在定位精度及状态判断上具有优势。
To solve the problem of eye localization of human face in multi-angle state of color image,a new method for the location and extraction of the eye image based on color information is proposed,and the method is not influenced by head posture.Firstly,face is detected in color images.Next,the non-skin region in the face area is found based on the difference between skin and eyes/eyebrows.In order to remove the interference points,mathematical calculation is exploited for the non-skin part.Then,eyes are detected with the method of average complexity.The 2-D correlation coefficient is used as a symmetry similarity measure to check the factuality of the 2 detected eyes.Finally,the white and iris of the eyes are precisely located in the eye region,and the eye opening extent can be measured by calculating the eye area.The experimental results demonstrate the proposed method can accurately detect eyes and determine the state of the eye quickly.The method is simple and fast,and has advantage in positioning accuracy and state analysis.
出处
《重庆大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2010年第11期20-26,共7页
Journal of Chongqing University
基金
国家自然科学基金资助项目(30870652C100401)
山东省中青年科学家科研奖励基金资助项目(2007BS03010)
关键词
多角度
人脸肤色
特征提取
图像分析
计算复杂度
multi-angle
face color
feature extraction
image analysis
computational complexity
