An ADMM Algorithm for MLS-TV Based Image Denoising
- 주제(키워드) Image Denoising , ADMM Algorithm , Moving Least Squares
- 발행기관 고려대학교 대학원
- 지도교수 이연주
- 발행년도 2018
- 학위수여년월 2018. 8
- 유형 Text
- 학위구분 석사
- 학과 대학원 응용수학과
- 원문페이지 33 p
- 실제URI http://www.dcollection.net/handler/korea/000000081131
- UCI I804:11009-000000081131
- DOI 10.23186/korea.000000081131.11009.0000817
- 본문언어 영어
- 제출원본 000045953966
초록/요약
Denoising is the problem of removing the inherent noise from an image. We consider the standard noise model with additive white Gaussian noise. The total variation (TV) regularization problem is a popular method that has applications in image denoising, image deblurring, and image reconstruction. In this paper, we propose an algorithm for denoising which solves the moving least squares (MLS) model with total variation minimizing regularization term by using the alternating direction method of multipliers (ADMM). Extensive experimental results show that our scheme needs even less iterations than MLS-TV to obtain the results.
more목차
1 INTRODUCTION 3
2 RELATED WORKS 4
2.1 Split Bregman Iterative Method 4
2.2 The Alternating Direction Method of Multipliers 6
2.2.1 Stopping Criteria 10
2.3 The Moving Least Squares approximation 10
2.4 The MLS with TV Minimizing Model 11
3 The Proposed Algorithm for TV problems 13
3.1 The Details of Proposed Algorithm 13
4 Experimental Results 16
5 Conclusion 17
