针对在结构性噪声较严重的情况下，常规几何活动轮廓模型无法获得理想分割效果的问题，提出一种基于几何活动轮廓模型的人脸轮廓提取方法，该方法首先将人脸形状的椭圆性约束作为算子嵌入到几何活动轮廓模型中，并利用几何活动轮廓模型提取任意轮廓的优势来快速抽取出图象中类似椭圆的目标边缘 ;然后根据图象中人脸的先验知识，通过对检测到的椭圆目标进行进一步验证来找出最终人脸轮廓.由于采用变分水平集方法做数值计算，因此该方法不仅能够自然地处理曲线的拓扑变化和能较精确地提取出图象中的人脸轮廓，而且同时可以给出人脸水平旋转的大致角度等信息.实验结果表明，该方法是有效的.

Face contour extraction is important in facial feature extraction and in model-based coding. For the face boundary, classical edge detection techniques will fail to exploit the inherent continuity of face boundaries. As the shape of face boundary is not uniform and exhibit low overall curvature, geometric active contours are an attractive choice for the extraction of face boundaries, but the original geometric active contour model still has no way to characterize the global face shape. In this paper a new method that incorporates prior shape information into geometric active contours for face contour extraction is proposed. As in general a human face can be treated as an ellipse with a little shape variation, the prior face shape is represented as an elliptical curve. By combining the prior face shape with the geometric active model proposed by Chan and Vese, our improved geometric active contour model can capture face contour depending on both the image edges and the prior knowledge of face shape. Moreover, our model is implemented using variational level set approach, thus the transformation parameters (such as the rotation angle in plane) that maps the face boundary to the prior shape can be roughly estimated simultaneously. The experimental results show the efficiency and effectiveness of our method.