【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.更多还原

【Abstract】 In order to overcome expressions and posture changes in face recognition to improve the accuracy,a new part-based face recognition method is proposed.Image quality adjustment factor is introduced.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.更多还原