计算机工程与应用 ›› 2011, Vol. 47 ›› Issue (15): 191-193.

• 图形、图像、模式识别 • 上一篇    下一篇

法曲率驱动曲面演化图像去噪模型研究

朱立新,罗子娟   

  1. 中国电子科技集团公司 第二十八研究所 C4ISR技术国防科技重点实验室,南京 210007
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2011-05-21 发布日期:2011-05-21

Study on normal curvature driven surface evolution image denoising model

ZHU Lixin,LUO Zijuan   

  1. National Key Laboratory of Science and Technology on C4ISR,CETC 28th Research Institute,Nanjing 210007,China
  • Received:1900-01-01 Revised:1900-01-01 Online:2011-05-21 Published:2011-05-21

摘要: 曲面演化图像去噪模型将图像看作嵌入高维空间中的曲面,并利用曲面局部几何特征来区分图像边缘和噪声信息,以此控制图像曲面演化速率,取得了良好的处理效果。通过分析现有基于主曲率的曲面演化图像去噪模型的优缺点,提出利用能够更好地刻画图像曲面几何结构的法曲率来控制曲面的演化过程。该模型将现有主曲率驱动的模型推广到了更一般的法曲率驱动的曲面演化模型。同时,由于法曲率在描述图像边缘、纹理、噪声等方面优于已有方法采用的主曲率,该模型在保持图像边缘、纹理等细节信息方面能够取得更好的性能。

关键词: 图像去噪, 法曲率, 曲面演化

Abstract: Images are considered as surfaces in high-dimensional space by surface evolution driven image denoising models.The local surface geometrical features are applied to differentiate between image edges and noises and used to control the speed of image surface evolution,which has produced good performance.In this paper,the advantages and disadvantages of principal curvature driven surface evolution models are analyzed and the normal curvature driven surface evolution model is proposed.The new model is the generalization of the existed curvature driven surface evolution models.Because the normal curvature has better performance over principal curvature in describing the geometric features of image edges,textures and noise,the new model is demonstrated better in the preservation of image edges and textures.

Key words: image denoising, normal curvature, surface evolution