Open AccessSingle-scale Residual Dense Dehazing Network
Author(s) -
Nian Wang,
Aihua Li,
Zhigao Cui,
Yanzhao Su,
Yunwei Lan
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1881/3/032008
Subject(s) - computer science , residual , block (permutation group theory) , concatenation (mathematics) , feature (linguistics) , artificial intelligence , translation (biology) , preprocessor , image translation , layer (electronics) , scale (ratio) , stack (abstract data type) , image (mathematics) , computer vision , pattern recognition (psychology) , channel (broadcasting) , algorithm , computer network , materials science , mathematics , philosophy , linguistics , chemistry , composite material , biochemistry , geometry , quantum mechanics , programming language , physics , combinatorics , messenger rna , gene
Recently, image dehazing algorithm has been widely used in the preprocessing of target tracking and pattern recognition. A large number of end-to-end convolutional networks have achieved good results in dehazing by image translation. In this paper, we use Residual Dense structure, which has been prove effective in high resolution reconstruction, to build feature extract block, and stack these block to form a single-scale dehazing network. In order to further enhance the performance, we convey the feature of shallow layer to deep layer by channel concatenation. The results show that our network has achieved good results in both the synthetic haze removal and the real haze removal.