一种采用背景统计技术的视频对象分割算法  

A video object segmentation algorithm with background statistical technology

作　　者：王成儒[1] 顾广华[1] 

机构地区：[1]燕山大学电子与通信工程系,河北秦皇岛066004

出　　处：《光电工程》2004年第8期57-60,共4页Opto-Electronic Engineering

摘　　要：利用背景统计技术从累积的帧差信息中构建出完整、可靠的背景区域,并将其与当前帧相比较,得到初始对象分割掩膜;再对之进行后处理,以消除噪声影响和平滑对象边界,从而获得较好的对象分割掩膜,并提取出视频对象。该算法不需要预知运动对象的形状、数目等,就能较好地从静止背景中分离出目标,实验证明,它具有一定的实用性和鲁棒性。Complete and reliable background region can be constructed from the accumulated frame difference information by using background statistical technology. As compared with current frame, we obtained the initial object segmentation mask. Then a post-processing is applied to the obtained object segmentation mask to remove noise and smooth the object boundary, thus we can obtain a better object segmentation mask and extract video object. The algorithm does not need to know the shape and number of motion objects beforehand, and it can separate objects from stationary background. The experiments show that this algorithm has certain practicality and robustness.

关 键 词：图像分割 视频图像 帧间差分 背景统计 目标测量 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TN941[自动化与计算机技术—计算机科学与技术]