基于偏移场模型空变失真图像序列的视频稳定化技术

Video stabilization technique for spatially variable distorted image sequence based on dynamic displacement field model

基于动态偏移场模型,提出了一种适用于空变失真图像序列的视频稳定化技术.以空不变图像序列的运动滤波技术为基础,采用空变图像序列的运动估计和频域滤波技术,实现了透视失真序列的稳定化.实验结果表明,经稳定化处理后的透视失真视频序列,高频抖动被滤除,图像序列呈现出稳定的透视失真现象.如果已知失真的物理机理,实现稳定化的同时还可以实现透视失真的校正.该技术不仅可以应用在视频稳定化系统中,还可以应用到其他空变失真图像序列的失真校正领域.

In this paper, a video stabilization technique based on Dynamic Displacement Field （DDF） model is introduced to stabilize the spatially variable distorted image sequence. Based on motion filtering technique of the spatially invariable image sequence, motion estimation and frequency filtering techniques of the spatially variable image sequence are used to stabilize the projective distorted image sequence. Experimental results show that the proposed method removes high-frequency jitter and stabilize the projective distortion after motion filtering. If we know the physical factors that create the distortion, we can correct the projective distortion while stabilizing the projective distortion. It can not only be potentially applied in to video stabilization system but also be applied in to other practical spatially variable distortion correction fields.