DOI:10.16136/j.joel.2023.02.0219基于天空区域分割与置信度图导向融合的去雾方法研究孙开伟*,冉雪,李彦,宣立德(重庆邮电大学数据工程与可视计算重点实验室,重庆400065)摘要:基于暗通道先验的去雾算法总是存在复原结果中天空区域处理不佳等问题,为了进一步优化对传输函数的估计,本文提出一种基于置信度图导向融合的传输函数优化方法。首先,将雾天图像的天空区域分离出来,以达到对天空区域的优化;计算窗口级暗通道与像素级暗通道,以平滑传输函数在物体边缘并保留小于窗口尺寸的细节特征;最后,计算窗口级暗通道与像素级暗通道之间的置信度图,以其为导向对两者进行融合得到优化的传输函数图,实现图像去雾。实验结果表明,本文算法可达到很好的复原结果优化效果。关键词:图像去雾;暗通道先验;天空区域分割;置信度导向融合中图分类号:TP391文献标识码:A文章编号:1005-0086(2023)02-0147-09Dehazingalgorithmbasedonskyregionsegmentationandreliabili-tymapguidedfusionSUNKaiwei*,RANXue,LIYan,XUANLide(KeyLaboratoryofDataEngineeringandVisualComputing,ChongqingUniversityofPostsandTelecommunica-tions,Chongqing400065,China)Abstract:Theskyregionintherestorationresultsfordehazingalgorithmbasedondarkchannelprioral-waysexiststhedrawbackssuchashalos.Inordertofurtheroptimizetheestimationoftransmissionfunction,thispaperproposesatransmissionfunctionoptimizationmethodbasedonreliabilitymapguidedfusion.Firstly,theskyregionofthehazyimageissegmentedandoptimized;thewindowleveldarklevelchannelandpixeldarklevelchannelarecalculatedtoensurethesmoothnessofthetransmissionfunctionattheedgeoftheobjectsandtheoutstandingofthedetailfeaturessmallerthanthewindowsize;finally,thereliabilitymapbetweenwindowleveldarkchannelandpixelleveldarkchanneliscalculated.There-finetransmissionfunctionmapisobtainedtorealizeimagedehazingbythefusionofthesetwodarkchannelsthroughtheguidanceofthereliabilitymap.Theexperimentalresultsshowthattheestimationoftransmissionfunctionoptimizedbytheproposedmethodcouldachievefineimagedehazingeffects.Keywords:imagedehazing;darkchannelprior;skyregionsegmentation;reliabilityguidedfusion0引言水滴、灰尘以及其他空气中的悬浮粒子,统称为气溶胶,对进入成像系统的光进...
