视频前景提取算法研究

Research on Video Foreground Extraction Algorithm

为了改善常用视频前景提取算法提取到的前景目标存在噪点、空洞,以及边缘不够平滑等缺点,提出对提取到的前景目标图片先进行中值滤波去除噪声与背景波动的干扰,再对图片进行开运算,以减少目标空洞并平滑目标边缘的改进方法。在静态背景下选用单高斯模型和RPCA方法、在动态背景下选用混合高斯模型方法,并加上提出的改进方法进行对比试验。实验结果表明,提出的方法能很好地解决常用算法存在的上述问题,充分证明了该方法的有效性。

In order to improve the common video foreground extraction algorithm to extract the prospects of the target there is noise,empty,the edge is not smooth enough shortcomings.We proposed to extract the foreground target image first median filter to remove noise and background fluctuations,and then open the image to reduce the target hole and smooth the target edge of the improved method.The experimental results show that we choose the single Gaussian model and the RPCA method in the static background,and use the mixed Gaussian model method in the dynamic background.The experiment is carried out with the improved method.The noise reduction is reduced,and relatively smooth.The conclusion is that the proposed method can solve the above problems of the commonly used algorithm and prove the validity of the method.