多角度人脸最佳视角选取方法

摘要

Aiming at some limits for selecting an optimal view from multiple views existing in the literature such as　complex computation,low robustness,an approach to choosing an optimal facial view is developed based on facial view　evaluation. YCbCr and YCgCr are adopted to detect face according to human face skin with some better clutter characteristics. The location of the lip is determined through the color difference inherent in face and lip. The facial view is　evaluated by the difference between the facial region and the lip region. An optimal facial view is chosen from multiple　views according to the facial view evaluation coefficient. Experimental results show that the proposal is efficient.%针对目前多视角人脸最佳视角方法计算复杂、鲁棒性低,提出了一种基于人脸视角评价因子的人脸最佳视角自动选取方法.根据人脸肤色具有良好的聚类特性,结合YCbCr和YCgCr色彩空间进行人脸检测,利用嘴唇色彩与人脸肤色的差异性,确定嘴唇区域位置.由人脸区域与嘴唇区域进行人脸视角评价,从多目摄像机系统中自动选择人脸最佳视角.实验表明所提方法有效、可行.