View : 1005 Download: 0

Full metadata record

DC Field Value Language
dc.contributor.advisor윤정호-
dc.contributor.author전세희-
dc.creator전세희-
dc.date.accessioned2020-02-03T16:32:39Z-
dc.date.available2020-02-03T16:32:39Z-
dc.date.issued2020-
dc.identifier.otherOAK-000000163170-
dc.identifier.urihttp://dcollection.ewha.ac.kr/common/orgView/000000163170en_US
dc.identifier.urihttps://dspace.ewha.ac.kr/handle/2015.oak/253127-
dc.description.abstractThe objective of this study is to develop an algorithm to remove Poisson-Gaussian noise. The algorithm is implemented in three-steps. First, we exploit the Generalized Anscombe Transformation (GAT) [4], which is one of the Variance Stabilization Transformations. Applying the GAT algorithm, we can transforms a Gaussian-Poisson noise to a Gaussian noise with unity variance in the entire image. As a second step, after GAT, we apply the technique of Improved Weighted Nuclear Norm Minimization [8] to remove the transformed noise which is treated as Gaussian noise. The motivation of using ImpWNNM is to use the gradient information when finding locally similar patches, such that it yields improved results. Also, we maintain the image structure by using Laplacian values. By doing so, the edge information is well preserved in the denosed images. After applying the Improved WNNM as Gaussian filter, we use an exact unbiased inverse [7] as a last step. Some numerical results are demonstrated to display the perfomance of the proposed denoising algorithm for Poisson-Gaussian noise. ;이 논문에서는 Generalized Anscombe Transformation과 WNNM을 사용하여 Poisson-Gaussian Noise가 있는 이미지 denoising 방법을 소개한다. 보통, Poisson-Gaussian noise를 제거하기 위해 Variance Stabilization Transformation 중 하나인 Generalized Anscombe Transformation(GAT)을 사용하는데, GAT후의 이미지를 Poisson noise가 아닌 전체 이미지에서 하나의 Variance를 가진 Gaussian noise를 가졌다고 가정하고 Gaussian denoising filter 중 하나인 WNNM을 개선한 Improved WNNM을 이용하여 노이즈를 제거하고 다시 Inverse GAT를 적용한다. Gaussian filter로 Improved WNNM을 사용할 때 Edge의 손실을 방지하기 위해 gradient와 Laplacian 정보를 이용하여 similar patch를 찾는 방법을 사용해 성능을 높인다. 흔히들 approximation에서 사용하는 Moving Least Squares(MLS)을 이용하여 정보를 얻는다. 그럼으로써, edge 정보를 더 잘 유지시키며 노이즈도 제거하는 효과를 얻을 수 있다. 여러가지 테스트 이미지에 이 알고리즘을 적용하여 Poisson-Gaussian noise를 제거하는 결과를 확인할 수 있다.-
dc.description.tableofcontents1 Introduction 1 2 Related Work 3 2.1 Generalized Anscombe Transformation(GAT) 3 2.2 Weighted Nuclear Norm Minimization 5 2.3 Inverse Generalized Anscombe Transformation 9 2.3.1 Algebraic Inverse 9 2.3.2 Asymptotically unbiased Inverse 9 2.3.3 Exact unbiased Inverse 9 2.3.4 Closed-form Inverse 10 2.4 Approximating Derivatives using Moving Least Squares 14 3 Improving WNNM 16 3.1 Finding more similar patches using Gradient and Laplacian information 16 3.2 Modified high-pass filtering 18 4 Experimental Results 21 5 Conclusion 34 References 35 국문초록 37-
dc.formatapplication/pdf-
dc.format.extent6418162 bytes-
dc.languageeng-
dc.publisher이화여자대학교 대학원-
dc.subject.ddc500-
dc.titleImproved Poisson-Gaussian Denoising Method Based on Weighted Nuclear Norm Minimization-
dc.typeMaster's Thesis-
dc.format.pageii, 37 p.-
dc.identifier.thesisdegreeMaster-
dc.identifier.major대학원 수학과-
dc.date.awarded2020. 2-
Appears in Collections:
일반대학원 > 수학과 > Theses_Master
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML


qrcode

BROWSE