摘要： 针对各种滤波方法对光流场进行去噪处理时会造成目标运动矢量改变的不足，提出一种基于奇异点检测的光流场去噪方法，利用表征信号奇异点特征的Lip指数建立噪声矢量判定准则，可以在目标无损的情况下检测出光流场噪声并消除。实验证明该方法能较好地消除噪声，同时不会造成目标运动矢量的损失。

关键词: 图像序列, 运动矢量, 光流场, 奇异点检测

Abstract: Conventional filtering methodologies may lead to insufficient changes of objective motion vector when it is used in noise-elimination. This paper proposes a novel noise-eliminating method in motion field based on singularity detection. It utilizes the Lip exponent to provide signal’s singularity characteristics and establish noise-vector judgment principle, which can eliminate the noise without lossing of the object. Experimental results demonstrate that it is able to eliminate noise perfectly without loss of the object.

Key words: image sequence, motion vector, optical flow field, singularity detection

中图分类号: 

• TP391.41