zhaozunjin / RetinexDIP

The pytorch implementation of RetinexDIP, a unified zero-reference deep framework for low-light enhancement.

RetinexDIP

The pytorch implementation of RetinexDIP: A Unified Deep Framework for Low-light Image Enhancement.

Z. Zhao, B. Xiong, L. Wang, Q. Ou, L. Yu and F. Kuang, "RetinexDIP: A Unified Deep Framework for Low-light Image Enhancement," in IEEE Transactions on Circuits and Systems for Video Technology, doi: 10.1109/TCSVT.2021.3073371.

Install

• scipy==1.2.1
• numpy==1.19.4
• opencv-python==4.1.1
• Pillow==8.1.2
• torch==1.2.0
• torchvision==0.4.0

Files Structure

├─data │ ├─test ├─net ├─output │ ├─illumination │ └─reflection ├─result └─utils

Dataset

• DICM
• ExDark
• LIME
• NASA
• VV
• Fusion, X. Fu, D. Zeng, Y. Huang, Y. Liao, X. Ding, and J. Paisley, ”A fusion-based enhancing method for weakly illuminated images,” Signal Processing, vol. 129, pp. 82-96, 2016.
• NPE, S. Wang, J. Zheng, H. Hu, and B. Li, ”Naturalness preserved enhancement algorithm for non-uniform illumination images,” IEEE Transactions on Image Processing, vol. 22, pp. 3538-3548, 2013

Experiments

python Retinexdip.py --input data/test --result ./result

Before running the code, you must assure that every dataset is included in the input root directory. For example, these datasets should be included in the "./data/test":

datasets = ['DICM', 'ExDark', 'Fusion', 'LIME', 'NPEA', 'Nasa', 'VV']

Explanations for some hyperparameters:

• input_depth This value could affect the performance. 3 is ok for natural image, if your images are extremely dark, you may consider 8 for the value.

• flag

  This parameter from the function named $get_enhanced$ can be set as $True$ and $False$. If the input image is extremely dark, setting the flag as True can produce promising result.