Chinese Journal of Liquid Crystals and Displays, Volume. 39, Issue 7, 971(2024)
Fast and robust low-light image enhancement based on iterative propagation network
Fig. 2. （a）IPN framework and training process；（b）Architecture of
Fig. 3. Image enhancement performance test based on publicly available datasets.（a）~（c）Different samples：（1）low-light images，（2）~（9）enhanced images via BIMEF，VEViD，EnlightenGAN，MLLEN，RetinexNet，Zero-DCE，Kindred-Nets and IPN，（10）ideal-light images as ground truths；（d）Quantitative comparison results.
Fig. 4. Image enhancement performance test based on photographic images.（a）~（c）Different samples：（1）low-light image，（2）~（9）enhanced images via BIMEF，VEViD，EnlightenGAN，MLLEN，RetinexNet，Zero-DCE，Kindred-Nets and IPN.
Fig. 5. Image enhancement performance test based on fluorescence microscopy images.（a）~（c）Different samples：（1）Low-excitation fluorescence images，（2）~（9）enhanced images via BIMEF，VEViD，EnlightenGAN，MLLEN，RetinexNet，Zero-DCE，Kindred-Nets and IPN，（10）high-excitation fluorescence images as ground truths；（d）Quantitative comparison results.
Table 1. Model complexity and inference time of deep learning methods
View tableView in ArticleTable 1. Model complexity and inference time of deep learning methods
Method EnlightenGAN MLLEN RetinexNet Zero-DCE Kindred-Nets IPN Params/M 54.409 12.150 0.555 0.079 1.853 0.003 FLOPs/G 32.410 863.25 348.60 83.31 957.00 3.12 Time/s 1.781 4 0.4394 0.119 1 0.011 8 0.025 0 0.006 0
Get Citation
Copy Citation Text
Zhibo XIAO, Zhilong JIANG, Yan KONG. Fast and robust low-light image enhancement based on iterative propagation network[J]. Chinese Journal of Liquid Crystals and Displays, 2024, 39(7): 971
Category: Research Articles
Received: May. 29, 2023
Accepted: --
Published Online: Jul. 23, 2024
The Author Email: Yan KONG (ykong80@163.com)
