【摘要】 从视频序列中提取视频目标是 MPEG- 4中的关键技术之一。该文研究了从视频序列中自动地提取出运动目标的算法。该算法将运动检测和彩色分割相结合而得到具有精确边缘的运动目标。彩色分割采用了基于 RGB彩色空间的区域生长法。在运动检测中 ,用 Gauss模型描述噪声引起的像素值变化 ,利用 Gauss分布特性区分噪声和运动区域 ,从而得到完整的运动目标。实验结果表明所提出的算法能有效地提取出运动目标。更多还原

【Abstract】 A key technique in MPEG 4 is extracting video objects from video sequences. This paper focuses on automatic extraction of moving objects in video sequences. Color segmentation and change detection are combined to extract moving objects in video sequences with pixel wise precision. Region growing in RGB space is used for color segmentation. Statistic modeling is used for motion detection. The change caused by noise is described with a Gaussian model. Then the change caused by the motion is separated from the noisy background because it does not obey the Gaussian model. Simulation results show that the automatic object extraction algorithm presented is promising.更多还原