摘要： 通过分析基于边缘、区域分割和形变模型等3类轮廓提取算法，提出一种分别对三者进行优化综合的基于梯度矢量流-主动轮廓模型(GVF-Snake)的人体轮廓提取优化算法。利用内嵌置信度边缘检测得到理想边缘图，通过形态学方法提取物体粗略轮廓作为GVF-Snake模型的初始轮廓。采用优化GVF-Snake模型提取精确轮廓。将上述方法应用于人体图像轮廓提取，实验结果表明，该方法达到了预期目的且具有较好的实用性。

关键词: 轮廓提取, 主动轮廓模型, 梯度矢量流, 内嵌置信度边缘检测, 数学形态学

Abstract: A new approach of body contour extraction is raised in this paper. The ideal edge map is derived from edge detection with embedded confidence. Imprecise contour is extracted based on morphological image analysis and it is used as the GVF-Snake model’s initial contour. And an optimized GVF-Snake model is used to extract precise contour. The result from comparative experiment of extracting contour of human body shows that described approach has a good effect and practicability.

Key words: contour extraction, active contour model, gradient vector flow, edge detection with embedded confidence, mathematical morphology

中图分类号: 

• TP391