视频车辆监控系统图像抖动快速消除算法研究

Study on Fast Video Stabilization Algorithm for Vehicle Monitoring System

在开发智能视频车辆监控系统时,为满足图像抖动消除速度和效果的要求,通过分析系统特点提出了一种基于图像匹配的新算法.新算法利用在基准帧中预先划定的适合于进行图像匹配的区域,在当前帧中根据像素灰度的绝对差值和寻找其最佳匹配,通过预测的方法快速求出当前帧相对于基准帧的偏移量,从而用来消除抖动.实验表明,新算法在确保抖动消除效果的同时大幅度地提高了处理速度,满足了系统要求.

To meet the speed and effect requirement of video stabilization for a vehicle monitoring system in developing,a new fast algorithm based on image matching was proposed by analyzing system features.In the new algorithm,a predefined region that was suitable for image matching to calculate motion parameters was used by comparing its original position in base frame and its best matching position in current frame.The best matching position was found by searching the minimum sum of absolute difference of pixel＇s gray value,and forecasting method was used in the searching.Then the motion parameters were used for video stabilization.In experiments,the new algorithm demonstrated a very high speed along with a favourable effect,which met the requirements of the system.