为了能够较好地识别出人脸五官并将其准确定位,提出了一种基于形状识别的人脸轮廓线提取算法。该算法采用经典的Canny边缘检测算子提取图片边缘曲线,利用SVD分解估计边缘曲线各点的切线方向,由此估计出曲率。根据各点的切方向与曲率信息来判断边缘曲线的形状信息,识别出人脸的五官。实验结果表明,该算法能够较好地识别出人脸五官,得到被识别目标的形状信息。
To recognize the parts of humane face and locate them, an algorithm which can recognize the contours of humane face is proposed based on the shape information. The algorithm makes use of Canny edge detection operator to get edge curves of the picture, SVD decomposition is used to estimate the tangent directions of the points on the edge curves, after that, the curvatures are estimated by it. The shape information are gotten by the tangent direction and curvature of every point. Experiment has proved that this algorithm can recognize the parts of humane face accurately and the shape information as well.