一种提取抖动视频中前景目标的新方法

提取视频中的前景目标信息是视频处理领域非常重要的问题,考虑到现实生活中会出现监控摄像头不可避免地会出现晃动或偏移情况,造成监控视频短暂抖动,此时背景图像灰度和纹理信息都会受到较大的影响,从而给后期进一步分析前景信息带来了巨大的困难。为了兼顾纹理特征提取和噪声抑制两方面的要求,针对抖动视频的前景提取问题,提出了一种有效的融合小波变换和在线混合高斯模型的方案。首先运用仿射变换在线逐帧校准,接着利用小波变换对图像去噪,并建立自适应模型迭代上述过程,最后利用在线混合高斯模型提取前景。实验结果表明,与同类方法相比,该算法无论针对单目标还是多目标视频均可以有效去除抖动,得到较好的前景目标提取效果,具有较高的准确性和鲁棒性。

A New Approach for Extracting Foreground Targets in Jitter Video

Extracting foreground target information in video is an important issue in the field of video processing.In consideration of the fact that surveillance cameras inevitably shake or shift in real life,which causes brief jitter of surveillance video,the background image grayscale and texture information will be greatly affected,which brings great difficulties for further on.Aiming at the foreground extraction problem of jitter video,in order to meet the requirements of texture feature extraction and noise suppression,an effective scheme of fusing wavelet transform and online mixed Gaussian model is proposed.Firstly,the affine transformation is used to calibrate online frame-by-frame,then the image is denoised by wavelet transform,and an adaptive model to iterate the above process.Finally,the online mixed Gaussian model is used to extract the foreground.The experiment shows that compared with the similar methods,the proposed algorithm can effectively remove the jitter for both single-target and multi-target videos and obtain better foreground target extraction results with high accuracy and strong robustness.