Title

Remote Sensing Image Super-Resolution based on Siamese Networks with Cross Scale Non-Local Similarity for Segmentation of Buildings

Alternative Title

건물의 분할 영상을 위해 교차 크기의 비지역적 유사성을 활용하는 샴 네트워크 기반 항공 영상 초해상도 기법

DGIST Authors

Jaegeun Park ; Jae Youn Hwang ; Kyong Hwan Jin

Advisor

황재윤

Co-Advisor(s)

Kyong Hwan Jin

Issued Date

2023

Awarded Date

2023-02-01

Citation

Jaegeun Park. (2023). Remote Sensing Image Super-Resolution based on Siamese Networks with Cross Scale Non-Local Similarity for Segmentation of Buildings. doi: 10.22677/THESIS.200000655870

Type

Thesis

Description

Cross Scale Non-Local Similarity, Siamese Network, Super Resolution, Remote Sensing, Segmentation, 교차 크기의 비지역적 유사성, 샴 네트워크, 초해상도, 항공 영상, 분할 영상

Table Of Contents

Ⅰ. INTRODUCTION 1
Ⅱ. RELATED WORKS 6
2.1 Image Self-Similarity 6
2.2 Deep Learning-based Super Resolution 7
Ⅲ. METHODS 9
3.1 Overall Network Structure 9
3.2 Siamese Residual Non-local Similarity Module (SRNSM) 13
3.3 Siamese Cross Scale Non-Local Similarity Module (SCSNLS Module) 16
Ⅳ. RESULTS 19
4.1 Experimental Dataset and Evaluation 19
4.2 Training Details 21
4.3 Comparison between previous and SRNSM methods 22
4.4 Super Resolution results on the UCMerced Dataset 23
4.5 Super Resolution Results on the WHU Dataset 28
4.6 Segmentation Results on the WHU Dataset 31
Ⅴ. DISCUSSION 34
Ⅵ. CONCLUSION 40
REFERENCES 42
국문요약문 47

URI

http://hdl.handle.net/20.500.11750/45753
http://dgist.dcollection.net/common/orgView/200000655870

DOI

10.22677/THESIS.200000655870

Degree

Master

Department

Department of Electrical Engineering and Computer Science

Publisher

DGIST

Show Full Item Record