DeepVariationalFrameworkforSingleImageDehazing
- Title
- DeepVariationalFrameworkforSingleImageDehazing
- Author
- 임은우
- Advisor(s)
- 김태현
- Issue Date
- 2023.8
- Publisher
- 한양대학교
- Degree
- Master
- Abstract
- Recentresearchhasoftenreliedontherepresentationpowerofneuralnetworksandoverlookedseveralfactorsinvolvedinhazedegradation,includingtransmissionandatmosphericlight.Ingeneral,thesefactorsareunknown,resultingininherentuncertainties.Inthisstudy,weintroduceavariationalBayesianframeworkforsingleimagedehazingtoaddresstheseuncertaintiesandaccountforthefactorsinvolvedinhazedegradation.Consideringkeycomponentsofhazedegradationaslatentvariables,theposteriordistributionsareparameterizedbycorrespondingtwobranchesofneuralnetworks,respectively.Basedontheatmosphericscatteringmodel,theproposedframeworkresultsinanewobjectivefunctionthatenablescooperationbetweenthesebranchesbyjointoptimizationandleadstoanamplificationoftheperformanceofeachother.Furthermore,themodel-agnosticframeworkfacilitatesnotonlyeasyadaptationofanyexistingdehazingnetworkswithoutmodificationofarchitecturebutalsonoextraoverheadintheinferencephase.Extensiveexperimentshavedemonstratedconsistentimprovementsintheperformanceofbaselinemethodsacrossdifferentdatasetsandmodels.
- URI
- http://hanyang.dcollection.net/common/orgView/200000684623https://repository.hanyang.ac.kr/handle/20.500.11754/188217
- Appears in Collections:
- GRADUATE SCHOOL[S](대학원) > ARTIFICIAL INTELLIGENCE(인공지능학과) > Theses(Master)
- Files in This Item:
There are no files associated with this item.
- Export
- RIS (EndNote)
- XLS (Excel)
- XML