Publications

Detailed Information

Directional Wavelets for Scattered Data and Their Applications : 산재한 자료에 대한 방향성 웨이블릿 및 응용

DC Field Value Language
dc.contributor.advisor오희석-
dc.contributor.author장동익-
dc.date.accessioned2017-07-14T00:30:25Z-
dc.date.available2017-07-14T00:30:25Z-
dc.date.issued2012-08-
dc.identifier.other000000004480-
dc.identifier.urihttps://hdl.handle.net/10371/121135-
dc.description학위논문 (박사)-- 서울대학교 대학원 : 통계학과, 2012. 8. 오희석.-
dc.description.abstract이 논문에서는 자료의 방향을 잘 표현하는 방향성 웨이블릿을 이용하여 산재한 자료를 효과적으로 분석하기 위한 다중척도 방법을 제시한다. 고전적 웨이블릿과 방향성 웨이블릿은 정규 격자 형태의 자료만 분석할 수 있기 때문에 공간 자료처럼 비정규 격자에서 관측된 자료를 효과적으로 분석하지 못하는 한계가 있었다. 따라서 본 연구에서는 이러한 제약을 극복할 수 있는 새로은 방향성 웨이블릿을 제안한다. 이미지 안에 있는 모든 부드러운 곡선들은 아주 미세한 수준에서는 작은 방향성 기저함수를 이용하여 표현될 수 있기 때문에, 국부적인 선형 기저함수를 점진적으로 결합하여 자료의 비선형 구조를 효과적으로 표현할 수 다중척도 방법을 제시한다. 더욱이, 본 논문에서는 자료에 국부적 특성에 따라 방향이 변화하는 이방성 기저와 자료의 전반적인 곡선을 표현하기 위한 비선형 기저 함수를 고안하였다. 그리고 이를 이용한 자료의 다중해상도분석을 위하여 새로운 방향성 웨이블릿을 제시하였다. 또한 제안된 방법을 구현하기위한 효과적인 분해와 복원 알고리즘을 제시한다. 마지막으로, 간단한 예제를 이용하여 제안된 방법을 설명하고 모의실험을 통하여 제안된 방법의 경험적인 성능을 다른 방법과 비교하였다.-
dc.description.abstractThis research introduces a multiscale method of constructing directional wavelets for the representation and analysis of scattered data. The proposed method overcomes limitations of classical wavelets and directional wavelets defined for regularly spaced observations, which are not able to capture directional structures of scattered data such as spatial observations. The central motivation of the proposed method is based on the fact that a smooth curve can be represented with linear directional bases in fine resolution levels. By the progressive merging of local linear directional bases, we effectively approximate the whole nonlinear directional data structure. We construct a directional nested network design with the progressive splatting that accounts for local directions. Furthermore, a data-adaptive anisotropic basis for the local directional feature and nonlinear basis for the global curve direction are proposed. Then we define the multiscale directional basis representation in multiresolution analysis. We also present efficient decomposition and reconstruction algorithms for the proposed directional wavelets. Finally, the implementation of the proposed method for scattered data is described, and through simulation studies the empirical performance of the proposed method is compared with other methods.-
dc.description.tableofcontents1 Introduction --- 1
2 Classical Wavelets and Directional Wavelets for Regular Data --- 4
2.1 Classical wavelets --- 5
2.1.1 One-dimensional wavelets --- 5
2.1.2 Multiresolution analysis (MRA) --- 8
2.1.3 Two-dimensional wavelets --- 12
2.2 Directional wavelets for regular data --- 21
2.2.1 Ridgelets --- 23
2.2.2 Curvelets --- 26
3 Wavelets for Scattered Data --- 30
3.1 Background --- 30
3.2 Second-generation wavelets --- 33
3.2.1 The lifting scheme --- 33
3.2.2 Lifting in two dimensions --- 35
3.3 Wavelets based on nested network design --- 38
3.3.1 Multiresolution analysis from bottom-up design --- 38
4 Directional Wavelets for Scattered Data --- 49
4.1 Directional nested network design --- 50
4.1.1 Data-adaptive local direction estimation --- 50
4.1.2 Progressive splatting --- 53
4.1.3 Directional basis function --- 58
4.1.4 Nonlinear direction estimation --- 67
4.2 Multiscale representation with directional wavelets --- 78
4.2.1 Multiscale directional basis function representation --- 78
4.2.2 Multiresolution analysis --- 79
4.2.3 Multiscale directional wavelets for scattered data --- 82
4.2.4 An example of multiscale direction estimation --- 83
4.3 Wavelet shrinkage --- 90
4.3.1 Thresholding --- 90
4.3.2 Level-dependent thresholding --- 92
5 Simulation Studies --- 95
5.1 Numerical examples --- 95
5.1.1 Test functions --- 96
5.1.2 An implementation of multiscale direction estimation --- 97
5.1.3 Comparison with classical wavelets and thinplate smoothing spline --- 102
6 Conclusion --- 106
6.1 Future research --- 107
-
dc.formatapplication/pdf-
dc.format.extent31451155 bytes-
dc.format.mediumapplication/pdf-
dc.language.isoen-
dc.publisher서울대학교 대학원-
dc.subjectDirectional wavelets-
dc.subjectDirectional nested networks-
dc.subjectMultiresolution analysis-
dc.subjectNonparametric function estimation-
dc.subjectScattered data-
dc.titleDirectional Wavelets for Scattered Data and Their Applications-
dc.title.alternative산재한 자료에 대한 방향성 웨이블릿 및 응용-
dc.typeThesis-
dc.contributor.AlternativeAuthorDongik Jang-
dc.description.degreeDoctor-
dc.citation.pagesvii, 117-
dc.contributor.affiliation자연과학대학 통계학과-
dc.date.awarded2012-08-
Appears in Collections:
Files in This Item:

Altmetrics

Item View & Download Count

  • mendeley

Items in S-Space are protected by copyright, with all rights reserved, unless otherwise indicated.

Share