摘要
主要介绍了光流的理论框架及应用背景,描述了主流的光流算法及逐步演变的数学求解模型.重点从保边平滑、抗光照变化影响、大位移光流、异质点滤除、实时性计算等5个不同的角度全面分析了光流模型优化求解中的技术性重点及难点问题,详细介绍了已有的解决策略并对今后的研究方向进行了展望,指出大位移光流、抗光照变化影响、实时性计算等方向将成为将来的研究热点.
This article primarily addressed the theoretical framework and the application background of optical flow estimation technology, and described the state-of-the-art popular optical flow algorithms and the evolving mathematical models. The technical emphasis and difficulties involved in the optimization of classical optical flow energy model were analyzed from 5 different aspects: edge preserving smoothing, vary illumination tackling, large displacement optical flow estimating, outlier removing and real-time computing. The existing proposed corresponding solutions were introduced in detail and the substantial research directions in the future were prospected. This article conclude that large displacement optical flow estimating, vary illumination tackling, and real-time computing will become the attractive topics on optical flow in the future.
出处
《北京工业大学学报》
CAS
CSCD
北大核心
2013年第11期1638-1643,共6页
Journal of Beijing University of Technology
基金
国家自然科学基金资助项目(61105033)
北京市自然科学基金资助项目(KZ201110005004)
关键词
光流场
变分模型
正则项
数据保真项
原始对偶算法
optical flow field
variational model
regularization term
data fidelity term
primal dual algorithm