摘要
提出采用基于肤色特征的最短生成树方法从复杂景物中对人脸进行分割，利用色调和饱和度空间进行联合分割．其方法优于在灰度空间中进行的分割，并在一定程度上不受阴影和光强的影响．最短生成树方法考虑相邻象素的相互关系，可提高分割的鲁棒性和准确性．实验证明，本方法有一定实用价值．
An approach using the shortest spanning tree based on skin color feature to segment and locate human face from complex scenes is proposed. It uses both hue space and chromatic space to do the segmentation, which is better than the segmentation done in gray space, and it will not be affected by highlight and shadow to a certain extent. The shortest spanning tree approach considers the relationship of the neighbouring pixels, thus improves the robustness and accuracy of the segmentation. It is proved from experiments that this approach is of certain practical value.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
1998年第1期12-17,共6页
Journal of Shanghai Jiaotong University
基金
国家自然科学基金
关键词
颜色分割
最短生成树
人脸分割
肤色
计算机视觉
color segmentation
shortest spanning tree
human face segmentation
color space
