«上一篇/Previous Article|本期目录/Table of Contents|下一篇/Next Article»

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

27

期数:

2017年03期

页码:

97-102

栏目:

智能、算法、系统工程

出版日期:

2017-03-10

文章信息/Info

Title:

 Image Rectification Method Based on Multiple Low-rank Textures Extraction

文章编号:

1673-629X（2017)03-0097-06

作者:

 马金辰；谢世朋；李海波

 南京邮电大学 通信与信息工程学院

Author(s):

 MA Jin-chen; XIE Shi-peng; LI Hai-bo

关键词:

 ; 低秩纹理; 增广拉格朗日乘数法; 图像分割; 多分辨率; 分支定界

Keywords:

 ; low-rank textures; augmented Lagrangian multiplier method; image segmentation; multi-resolution approach;  branch-and-bound scheme

分类号:

TP391.41

文献标志码:

A

摘要:

 低秩纹理是图像处理领域中具有重要几何意义的结构,通过提取低秩纹理可以对畸变或受损图像进行有效校正;但现有的基于低秩纹理提取的图像校正方法只能把感兴趣区域看作一个整体,从而导致许多复杂情况下的校正无法获得理想结果.针对上述问题,对现有算法进行改进,实现感兴趣区域的多个低秩纹理分别提取,用以解决实际应用中的多种畸变和受损图像的恢复问题.对选定区域进行分割,对每个子区域分别进行低秩校正,加强初始化窗口对纹理的适应性,从而优化校正效果.大量实验结果表明,该方法可以有效处理同一感兴趣区域中的多个低秩纹理,在凸平面、多个不兼容区域和复杂纹理等多种情况下仍然可以获得正确的处理结果,且校正效果与原算法相比更加符合人眼视觉特性.

Abstract:

 Low-rank textures play an important role in image processing fields. By extracting the low-rank textures accurately,the distort-ed or damaged image can be rectified effectively. However,the existing methods based on texture extraction always rectify the region of interest as a whole,which makes it comes to nothing in many complex cases. Aiming at the above-mentioned problems,an improved rec-tification algorithm is proposed based on multiple low-rank textures extraction. The new rectification algorithm is able to extract multiple low-rank textures respectively and solve a large number of problems of image renewing for distorted or damaged images in practical ap-plication. In order to optimize the experimental results,the selected regions are segmented and each sub-region is rectified respectively which can enhance the adaptability of the initialization window to the texture. Experimental results demonstrate that the proposed method is able to rectify multiple low-rank textures in the same region of interest. In complex circumstances such as convex plane,multiple in-compatible regions,complex textures and so on,the ideal processing results can be obtained with it. Further more,the results are more in accord with human vision characteristics,which is a critical evaluation criteria.

常用功能

更新日期/Last Update: 2017-05-18