摘要
针对基于像素的自适应分割检测算法在建立背景模型时容易产生鬼影的问题,根据背影视觉提取算法中相邻像素点拥有相近像素值的空间分布特性,将待定的前景像素值与邻域像素值的差值通过加权比较其与自适应阈值的大小,来确定该像素点是前景像素点还是鬼影像素点。若为鬼影像素点则判定为背景像素点,并更新其背景模型。通过对前景像素点的二次判断,达到迅速去除鬼影的目的。实验结果表明,改进后的算法相比于原算法能更快速地去除鬼影。
It is probable that ghost would be produced when background model is established with pixel-based adaptive segmenter algorithm. According to the theory that adjacent pixel points have similar pixel values in spatial distribution in VIBE algorithm, the difference of foreground pixels was computed which are judged in the fi rst time and the neighborhood pixel. Then the difference and adaptive threshold value were compared through weight to determine whether the pixel is foreground pixel point or ghost which will be judged as background pixel. At last the background models would be updated. Through twice judgments of the foreground pixels, the ghost would be removed quickly. The experiment results show that the improved algorithm removes the ghost faster than the original method.
出处
《集成技术》
2015年第2期50-56,共7页
Journal of Integration Technology
关键词
鬼影去除
像素自适应分割
背景建模
运动目标检测
ghost removal
pixel-based adaptive segmenter
background model
motion detection
