计算机工程 ›› 2010, Vol. 36 ›› Issue (23): 217-219.doi: 10.3969/j.issn.1000-3428.2010.23.073

• 多媒体技术及应用 • 上一篇    下一篇

一种新的视频图像分割算法

胡学刚a,胡文涛b   

  1. (重庆邮电大学 a.数理学院; b.计算机学院, 重庆 400065)
  • 出版日期:2010-12-05 发布日期:2010-12-14
  • 作者简介:胡学刚(1965-),男,教授、博士,主研方向:偏微分方程,图像处理;胡文涛,硕士研究生
  • 基金项目:
    国家自然科学基金资助项目(10771226);重庆市自然科学基金资助项目(CSTC,2007BB2450);重庆邮电大学自然科学基金资助项目(A200931)

New Video Image Segmentation Algorithm

HU Xueganga,HU Wentaob   

  1. (a.College of Mathematics and Physics; b.College of Computer, Chongqing University of Posts and Telecommunications, Chongqing 400065, China)
  • Online:2010-12-05 Published:2010-12-14

摘要: 针对传统水平集需要不断重新初始化,不能直接应用到视频图像分割中的问题,提出一种新的视频图像分割算法。该算法根据图像序列中相邻的2帧,利用帧差法进行前期处理,并经过二元休憩和数学形态学的处理得到一个运动目标的模板,将该模板作为新的初始水平集进行演化,可以得到图像序列中的运动目标。实验结果表明,该算法能够得到运动目标的准确位置,避免传统水平集需要不断重新初始化的问题,减少计算复杂度。

关键词: 图像分割, 形态学处理, ChanVese模型, 主动轮廓线模型

Abstract: Specific to the problem that traditional level set need constantly reinitialization and can not be directly applied to the video image segmentation problems, this paper presents a new video image segmentation algorithm. The new method uses the framedifference method based on the adjacent two images sequence to do pretreatment, and through morphological processing, a moving target template can be acquired. The template is used as the initial level set to go through new evolutions and to get the moving targets of the image sequence. Experimental results prove this method can get the exact location of moving targets, which avoids the constant reinitialization of the level set and reduces the computational complexity.

Key words: image segmentation, morphological processing, ChanVese model, active contour model

中图分类号: