Deep Trident Decomposition Network for Single License Plate Image Glare Removal

This repository provides the dataset and code for the following paper:

Paper: Deep Trident Decomposition Network for Single License Plate Image Glare Removal

Authors: Bo-Hao Chen, Shiting Ye, Jia-Li Yin, Hsiang-Yin Cheng, and Dewang Chen

in IEEE TITS.

Abstract

Deep convolutional neural networks have achieved state-of-the-art performance for the removal of atmospheric obscuration. However, most relevant studies have focused on eliminating the effects of atmospheric obscuration but not on the glare in images caused by reflected sunlight. On the basis of a glare image formation model, we propose a deep trident decomposition network with a large-scale sun glare image dataset for glare removal from single images. Specifically, the proposed network is designed and implemented with a trident decomposition module for decomposing an input glare image into occlusion, foreground, and coarse glare-free images by exploring background features from spatial locations. Moreover, a residual refinement module is adopted to refine the coarse glare-free image into fine glare-free image by learning the residuals from features of multiscale receptive field. The experimental results indicated that the proposed network significantly outperforms state-of-the-art atmospheric obscuration removal networks on the built dataset.

Requirements

Dependencies

• cuda 10.0
• cudnn 7.3.1
• Python 3.5
• numpy 1.15.0
• openCV 4.3.0.38
• keras 2.3.1
• tensorflow 1.14.0
• scikit-image 0.14.2
• pillow 6.2.1
• sewar 0.4.3
• segmentation-models 0.2.1

Dataset

The sun glare image dataset can be downloaded from here.

Model

Pre-trained models can be downloaded from google drive or baidu drive (password: qpuf).

It was tested and runs under the following OSs:

• Windows 10
• Ubuntu 16.04

Might work under others, but didn't get to test any other OSs just yet.

Getting Started:

Usage

• Training

$ python train.py --train_path=/your/trainset/path --vaild_path=/your/valset/path --save_model_dir=/save/model/path --load_pretrain=/pretrain/generator/model/path

• Testing

$ python test.py --test_path=/your/testset/path --load_pretrain=/pretrain/generator/model/path

Demo

To test this code

$ python test.py --test_path=./dataset/test/synthesized_glare_images --load_pretrain=./save_weight/model.h5

The test results will be saved in: ./test_result/

Results

The images were the recovered results using the methods of (a) Engin et al., (b) Li et al., (c) Qian et al., (d) Yang et al., (e) Wei et al., and (f) Zhang et al. as well as (g) our previous method and (h) the method proposed in this study. Please refer to our paper for more information.

License + Attribution

This code is licensed under CC BY-NC-SA 4.0. Commercial usage is not permitted. If you use this code in a scientific publication, please cite the following paper:

@ARTICLE{ChenTITS2021,  
 author={Chen, Bo-Hao and Ye, Shiting and Yin, Jia-Li and Cheng, Hsiang-Yin and Chen, Dewang}, 
 journal={IEEE Transactions on Intelligent Transportation Systems},  
 title={Deep Trident Decomposition Network for Single License Plate Image Glare Removal},  
 year={2022}, 
 volume={23}, 
 number={7}, 
 pages={6596-6607}, 
 doi={10.1109/TITS.2021.3058530}}