摘要
提出一种基于先验信息的运动物体分割与跟踪算法。该算法假设帧间运动物体颜色直方图和大小等信息是相对不变的,利用动态规划思想最小化当前帧目标物体颜色直方图和大小与已知目标物体颜色直方图和大小信息的差距,从而找到较准确的目标物体轮廓。与梯度向量流算法相比,该算法对初始轮廓的鲁棒性较好,能在复杂的背景中找到物体轮廓。
This paper proposes an object segmentation and tracking algorithm based on prior information. It is based on the assumption that the color histogram and area of tracked object are nearly stationary from frame to frame. It finds the final contour of tracked object by minimizing the difference of color histogram and area between current flame and previous flame in dynamic programming way. Compared to gradient vector flow algorithm, the algorithm can find better contour in complicated scene, and is robust to the initial contour.
出处
《计算机工程》
CAS
CSCD
北大核心
2009年第12期140-142,共3页
Computer Engineering
关键词
动态轮廓模型
颜色直方图
梯度向量流
active contour model
color histogram
gradient vector flow
