计算机工程与应用 ›› 2011, Vol. 47 ›› Issue (12): 196-199.

• 图形、图像、模式识别 • 上一篇    下一篇

用图割构造长景物图像

袁晓光,刘 红,程 鸿,韦 穗   

  1. 安徽大学 计算智能与信号处理教育部重点实验室,合肥 230039
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2011-04-21 发布日期:2011-04-21

Producing panoramas of long scenes with method of graph cut

YUAN Xiaoguang,LIU Hong,CHENG Hong,WEI Sui   

  1. Key Lab of Intelligence Computing and Signal Processing of the Ministry of Education,Anhui University,Hefei 230039,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2011-04-21 Published:2011-04-21

摘要: 多视点长景物图像指的是由一系列沿某一景物连续拍摄的图像拼接而成的长景物图像。提出了一种基于图割的由多视点构造长景物图像的新方法,该方法将由摄像机拍摄的原始图像投影到将要形成的长景物图像主平面上;再根据长景物图像形成的几个条件构建能量函数,将视点选择问题转变为能量函数最小化问题;通过Boykov等人提出的快速近似能量函数最小化的方法使其最小化,从而得到满足条件的长景物图像。实验结果显示,用该方法拼接而成的长景物图像在拼接处显得相当自然,真正做到了无缝光滑过渡。

关键词: 长景物图像, 主平面, 能量函数, 图割

Abstract: The multi-viewpoint panorama of long scenes is produced from a relatively sparse set of images captured with a camera that is moved along the scene.A novel method for producing panoramas of long scenes based on graph cuts is proposed.Source images are projected onto the picture surface.The energy function is built based on several conditions of the long scenes,and the problem of view selection is translated into the problem of minimizing energy function.The energy function is minimized using Boykov’s method.Experimental results show that the panoramas produced from this algorithm seems quite natural,and the transitions between different regions of linear perspective are more natural and seamless.

Key words: panoramas of long scenes, picture surface, energy function, graph cut