基于几何代数的彩色光流场计算  

Estimation of Color Optical Flow Based on Geometric Algebra

作　　者：李茂宽[1] 关键[1] 

机构地区：[1]海军航空工程学院,山东烟台264001

出　　处：《光学学报》2009年第10期2837-2841,共5页Acta Optica Sinica

基　　金：国家自然科学基金(60672140;60802088);教育部新世纪优秀人才支持计划(NCET-05-0912)资助课题

摘　　要：提出了一种新的基于几何代数的彩色光流场计算方法,从新的角度出发讨论了运用几何代数的概念来解决彩色图像序列的可行性和简便性。在几何代数域内对彩色图像序列用多重矢量表示,证明了彩色光流场分析时进行约束条件时的测量问题,通过多重矢量而非多通道的图像处理方法来表示图像连续特性,从而扩大了测量范围;然后依据物体色彩在运动过程中保持不变的原理,在几何代数域内推导了彩色图像序列的光流约束方程及其解法。结果表明,该算法能够显著提高各种情况下的彩色光流场计算能力。A new optical flow estimation method for color image sequence is proposed,from a new point of view,the probability and simplicity of using the concept of geometric algebra to analyse color optical flow is discussed in detail. Color image sequence is presented by multivectors in geometric algebra,it is proved that if meet the constrained conditions for color optical flow analysis,a multivector instead of multichannel image processing method can be used to represent the continued distribution of images to expand the measurement range. then a new optical flow constraint equation and its solution in geometric algebra domain was deduced based on color invariance under motion,The experimental results of several synthesized and natural image sequences show that this algorithm can improve the ability of color optical flow estimation in different conditions greatly.

关 键 词：机器视觉 几何代数 光流场 图像序列 偏微分方程 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]