
Single image dehazing based on hidden Markov random field and expectation–maximisation
Author(s) -
Kwon O.
Publication year - 2014
Publication title -
electronics letters
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.375
H-Index - 146
ISSN - 1350-911X
DOI - 10.1049/el.2014.2093
Subject(s) - markov random field , random field , markov chain , artificial intelligence , image (mathematics) , computer science , field (mathematics) , hidden markov model , variable order markov model , pattern recognition (psychology) , markov model , computer vision , mathematics , algorithm , statistics , image segmentation , machine learning , pure mathematics
Single image dehazing using an estimated transmission map for evaluation is proposed. When estimating the transmission map using the conventional dark channel prior (DCP) algorithm and then detecting the haze regions using a matting algorithm, the resulting transmission map invariably includes some block artefacts, as the DCP algorithm is based on patch‐based processing. Therefore, a modified transmission map is proposed based on the hidden Markov random field (HMRF) and expectation–maximisation (EM) algorithm. Experimental results confirm that the proposed algorithm is superior to conventional algorithms in image haze removal.