采用部件和图像质量调整因子的人脸识别方法  

Part-based Face Recognition Using Image Quality Adjustment Factor

作　　者：潘可[1] 赵松[1] 张志坚[1] 张培仁[1] 

机构地区：[1]中国科学技术大学自动化系,安徽合肥230026

出　　处：《小型微型计算机系统》2009年第7期1399-1403,共5页Journal of Chinese Computer Systems

摘　　要：为了更好克服人脸识别中的姿态和表情等的影响,提高人脸识别精度,提出一种新的采用部件和图像质量调整因子的人脸识别方法.人脸图像根据眼睛的坐标被归一化,并被划分为多个部件,划分出的各部件分别给出判决分数.根据各部件的基础权值及图像质量调整因子,各部件的判决结果被融合并给出最终结果.开集测试表明,该方法优于基于整脸的方法和采用部件的简单加权融合的方法.In order to overcome expressions face recognition method is proposed. Image and posture changes in face recognition quality adjustment factor is introduced. to improve the accuracy, a new part-based Face image is normalized to a fixed size according to the eye coordinates. Decomposition is employed to extract different facial parts from the face image. Each part gives out a score. The final score is reached by fusion all parts＇ score according to each part＇s weight and its image quality adjustment factor. Experiment shows that our method outperformed the holistic method and the part-based method with fixed weights.

关 键 词：人脸识别 采用部件 融合 动态权值 图像质量 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]