摘要
唇部提取是唇读系统和基于唇部的身份识别系统中的重要预处理过程,提取效果直接影响系统性能,现有的唇部提取方法对被提取人肤色有一定依赖性,且对带胡须和露齿图像的唇部提取效果较差。针对该问题,通过对常用色度空间各分量的可分离性进行研究,提出基于Lab色度空间的a分量唇部提取方法。实验结果证明,该方法能实现唇部的自动分割和提取,具有较好的鲁棒性。
For lip-reading system and authentication system based on lip, lip extraction is a very important preprocessing process. It affects the performance of application systems. The existing methods for lip extraction depend on the skin color of the people, and they are not effective for the image with tooth and brave. Aiming at this problem, based on the research of the diversity for different color space, this paper presents a lip extraction method using a component of Lab color space, experimental results show it is effective and robust.
出处
《计算机工程》
CAS
CSCD
北大核心
2011年第3期19-21,24,共4页
Computer Engineering
基金
国家自然科学基金资助项目(U0735004)
关键词
唇部提取
Lab色度空间
唇读
排红法
色度分布差异
lip extraction
Lab color space
lip reading
Red Exclusion(RE)
Chroma Distribution Diversity(CDD)