基于形变轮廓模型的视频图像中人物领域的追踪

Extraction of human region from image sequence based on deformation contour model

基于头部轮廓设置人物领域轮廓模型,建立形变轮廓模型模板,通过模板评价函数完成模板的更新修正,实现不基于帧间差分和移动向量亦能在视频图像中追踪移动对象。基于形变轮廓模型算法,可实现摄像头转动或背景变动,且在背景较复杂的情况下从视频图像中检测移动人物。实验结果验证了该算法的有效性与鲁棒性,可快速准确抽取视频图像中的人物领域。

This paper can get contour model of head by using face detection results. Based on the contour model of head and Gaussian function to found the upper part of the human body model （object mask）. In proposed method, the object mask is amend- ed in the human tracking process based on condensation method. Experiment results indicated that the proposed method is robust to complex background.