具有前景目标的动态场景视频快速稳像算法

Fast stabilization algorithm of dynamic scene video with multiple foreground objects

为了提高具有前景目标的动态场景视频的稳像效果,采用了一种基于块绝对差分的运动目标自动提取方法。在提取运动目标的基础上,提出通过自适应设定相关阈值的方法,来预判相关背景模块是否具有足够的信息,以减少用于运动矢量估计的模块数量。仿真实验结果表明,对于具有前景目标的动态场景视频,在减少运动目标提取时间的基础上,能够有效提高稳像算法的速度和精度,对实际视频取得了比较理想的稳像效果。

In order to improve the stabilization effect of dynamic scene video sequences with multiple moving foreground objects, a new method based on block absolute difference is proposed to extract the foreground objects automatically. Based on extracting the moving objects, the background blocks are selected to estimate the motion parameters with the defined threshold, so the background blocks for estimating motion parameters are greatly reduced. For moving scene with multiple moving foreground objects, due to the reduction of the time of extracting moving objects, the simulation results shown that the running time of the proposed method is reduced effectively, the stabilization effect is improved, and more ideal effect is realized compared with the traditional method.