CASIA OpenIR  > 毕业生  > 硕士学位论文
Alternative TitlePartial Differential Equation Application in Texture Analysis and Edge Detection
Thesis Advisor彭思龙
Degree Grantor中国科学院研究生院
Place of Conferral中国科学院自动化研究所
Degree Discipline模式识别与智能系统
Keyword偏微分方程 纹理元 亚像素 边缘检测 Partial Differential Equation Texton Subpixel Edge Extraction
Abstract偏微分方程在图像处理中有广泛的应用。本文在偏微分方程特性分析的基础上,从纹理分析和边缘检测两个方面着手,提出了应用于亚像素边缘检测的改进的水平集重构和用于纹理元构造的各项异性扩散滤波器。 基于纹理元的纹理分析常用算法在构造纹理特征时产生的冗余性使得检索过程非常耗时。针对检索时间损耗这一关键问题,本文提出一种基于各向异性扩散滤波的纹理元构造算法。由于各向异性扩散使得大量相近特征得到有效归并,本算法在保证纹理元构造可靠性的基础上极大节省了特征向量空间、纹理元构造以及纹理检索时间。本文以纹理检索为例,证实了所提算法在与传统Gabor滤波器达到相似的纹理元构造效果的基础上,将运算时间缩短到原来的53%。 常用的亚像素边缘定位算法采取局部表面估计或结构模型实现局部边缘的定位,不可避免由于离散化导致的锯齿效应和由于局部模型带来的不连续边缘定位。本文在扩散方程的基础上提出了一种改进的水平集重构算法, 通过改进的水平集重构实现边缘沿切线方向的平滑;同时加入角点定位判断,在角点处加入角点限制和拓扑限制以保证角点处的精确定位。本文通过几何图像和自然图像的亚像素边缘提取对比试验验证了所提算法能在保持亚像素精度的同时消除锯齿效应,保持边缘的平滑和连续性。
Other AbstractPartial differential equation has broad application in image processing. In this paper, we focus on its application in texture analysis and edge location. For that purpose, two new partial differential equation application are proposed. The first process is based on an anisotropic diffusion. Recently, texton has been an intensely researched topic in texture analysis field. Gabor filters based method is commonly applied in this field. But the redundancy of Gabor filters makes texton based methods high time consumption. To reduce the time consumption, anisotropic diffusion (AD) filters are introduced into feature extraction for texton construction. Because anisotropic diffusion use much less nonlinear features to get the feature information of the Gabor filters, this method largely reduces feature vectors and time consumption while getting reliable texton construction. Texture retrieval is taken as example to compare the difference in texton construction and time consumption between Gabor filters and anisotropic diffusion filters. The proposed method’s effectiveness is proved to save 53% time compared with Gabor filters. The second process is based on an improved level set reconstruction. Traditional subpixel edge location algorithms use local model to realize precise edge location. Zigzag effect and discontinuous edge location are side effects. We present a new subpixel edge location algorithm based on partial differential equation (PDE). Level-set reconstruction (LSR) is introduced to smooth edge along their tangents. Anchor and topology constraints are used to avoid over-smoothing and keep edge topology structures. Subpixel edge location experiments with geometric and natural image shows, this method can remove zigzag effect while keeping edge smoothness and continuity.
Other Identifier200228014603532
Document Type学位论文
Recommended Citation
GB/T 7714
易沫. 偏微分方程在纹理分析和边缘检测中的算法研究[D]. 中国科学院自动化研究所. 中国科学院研究生院,2005.
Files in This Item:
There are no files associated with this item.
Related Services
Recommend this item
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[易沫]'s Articles
Baidu academic
Similar articles in Baidu academic
[易沫]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[易沫]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.