검색 상세

An Annotation and Content-Based Leaf Image Retrieval System

  • 남윤영 (Nam, Yun-Young, 아주대학교 정보통신전문대학원)

초록/요약

For decades, many researchers have proposed and developed methods of content-based image retrieval using image features such as color, shape, texture, and spatial relationship. In particular, shape-based image retrieval has received considerable attention recently with interest in establishment of large biological databases. If images contain similar color or texture, shape-based image retrieval method is more effective than other methods using color or texture. For instance, leaves of most plants are green or brown; but the leaf shapes are distinctive and thus can be used for identification. In this thesis, we present an effective and robust leaf image retrieval system based on shape feature. For the effective measurement of leaf similarity, we consider contour-based shape feature as well as region-based venation feature which corresponds to the blood vessel of organisms. In the contour-based shape domain, a matrix of interest points is constructed to model the similarity between two leaf images. In order to improve the retrieval performance, an improved Minimum Perimeter Polygons (MPP) and an adaptive grid-based matching algorithm are developed for the shape representation and data matching, respectively. In the region-based venation domain, an adjacency matrix is constructed from intersection and end points of a venation to model the similarity between two leaf images. Additionally, we propose a novel leaf image classification scheme which first analyzes the venation for leaf categorization and then extracts and utilizes shape feature to find similar ones from the categorized group in the database. The venations are represented using points selected by the curvature scale scope corner detection method on the venation image, and categorized by calculating the density of feature points using non-parametric estimation density. In order to reduce the matching time, we have proposed an adaptive matching algorithm using leaf symmetric distribution. Using this property, matching scope on the shape can be reduced. We also implemented a grid-based matching algorithm on our system to fast access to records and efficient process range queries. Several tests have shown that the algorithms are suitable for image databases with a varying number of attributes. We implemented a prototype system and performed various experiments to show its effectiveness. Its performance is compared with other methods including Centroid Contour Distance (CCD), Fourier Descriptor, Curvature Scale Space Descriptor (CSSD), Moment Invariants, and MPP. Experimental results on one thousand leaf images show that our approach achieves a better performance than other methods

more

목차

Chapter 1 Introduction = 1
1.1 Content-based Image Retrieval = 2
1.2 Shape-based Image Retrieval = 3
1.3 Information on Leaf Image = 4
1.4 Thesis Outline = 5
Chapter 2 Literature Review = 7
2.1 Feature extraction = 7
2.1.1 Color = 8
2.1.2 Texture = 9
2.1.3 Shape = 11
2.2 Image Retrieval System = 13
2.2.1 QBIC = 14
2.2.2 Virage = 16
2.2.3 Photobook = 16
2.2.4 VisualSEEk and WebSEEk = 18
2.2.5 Netra = 19
2.2.6 STAR = 19
2.2.7 MARS = 20
2.2.8 SQUID = 21
2.2.9 ARTISAN = 22
2.2.10 Other systems = 22
Chapter 3 Background on Shape Representation and Matching = 24
3.1 Contour-based Shape Representation Techniques = 25
3.1.1 Fourier Descriptors = 26
3.1.2 Chain Codes = 26
3.1.3 Curvature Scale Space Representation = 28
3.1.4 Boundary Approximation = 31
3.1.5 Minimal Perimeter Polygon = 32
3.2 Region-based Shape Representation Techniques = 33
3.2.1 Common region-based methods = 34
3.2.2 Convex Hull = 35
3.2.3 Media Axis = 36
3.2.3 Grid Descriptors = 37
Chapter 4 Shape-based Leaf Image Retrieval System = 40
4.1 System Overview = 40
4.2 Image segmentation = 43
4.2.1 Contour-based Shape Extraction = 44
4.2.2 Contour-based Shape Representation = 44
4.3 Leaf Arrangement and Venation Representation = 49
4.4 Venation Classification = 53
4.4.1 Venation Types of Leaves = 55
4.4.2 Feature Points Extraction = 57
4.4.3 Feature Point Classification 60 =
4.4.4 Density Distribution of Feature Points = 63
4.4.5 Distribution of Parallel Venation = 64
4.4.6 Distribution of Palmate Venation = 66
4.4.7 Summary of Venation Classification = 68
4.5 Image Matching and Retrieval = 69
4.5.1 Basic Shape Matching = 69
4.5.2 Adaptive Shape Matching = 70
4.5.3 Grid-based Shape Matching = 72
Chapter 5 Results and Analysis = 76
5.1 Test Setup = 76
5.2 User Interface = 77
5.3 Retrieval Results = 78
5.3.1 Experiments on Contour-based Image Retrieval = 78
5.3.2 Experiments on Venation Classification = 80
5.3.3 Experiments on Response Time = 83
5.4 Chapter Summary = 87
Chapter 6 Conclusion = 88
6.1 Summary = 88
6.2 Future Research = 90
Appendix A A Representative Subset of Leaf Images = 91
Bibliography = 94

more
반출 Meta View 목록