Awesome

LIME: Low-light Image Enhancement

This is a MATLAB implementation of the paper, "LIME: Low-Light Image Enhancement via Illumination Map Estimation". It was done as a course project for Digital Image Processing (ECN-316), under the guidance of Prof. Saumik Bhattacharya.

The project report can be found here.
The paper can be found here.
The official website for the project can be found here. The demo software provided by the authors contains .p mat files, for which source code can't be read.

How to Use

Clone the repository:

git clone https://github.com/estija/LIME.git

Open MATLAB, go to the git repository folder.

Add paths

Run the following to the MATLAB command window:

addpath('./BM3D');
addpath('./imgs');

Load images and run

Run the following commands in the MATLAB command window:

img_in = imread('x.bmp'); 
[Ti, Tout, img_out, Iout] = lime_main_module(img_in, mu, rho, ds, ss, flag);

x is some image from imgs
flag = 1 to view the results.
Ti and Tout are initial and refined illumination maps, img_out and Iout are enhanced and denoised results.
Use the table mentioned below for selecting optimum values of mu, rho, ds, ss for each image.

Table for optimized hyperparameters

Name	`mu`	`rho`	`ds`	`ss`
building	0.01	1.188	10	1.5
cars	0.045	1.134	5	0.75
lamp	0.8	1.07	0.1	1
land	0.5	1.09	0.3	4
moon	0.01	1.2	1	0.5
paint	0.3	1.15	1	0.5
robot	0.01	1.25	10	1
table	0.002	1.035	100	1
wires	0.01	1.165	1	0.6

Information regarding other files:

lime_loop.m is the file for tuning the parameters for the solver, i. e. alpha, mu, rho.
lime_bf_loop.m is the file for tuning the parameters for the bilateral filter, i. e. ds, ss. By default, lime_bf_loop.m is used for post-processing.
histeq_all.m is the file for generating the results of applying histogram equalization, on the raw image, in five different ways.
denoise_bm3d.m is the file for denoising the enhanced result using BM3D.
The code for BM3D in file BM3D.m as well as other files related to it have been downloaded and used only for the purpose of comparison with bilateral filtering.

Results

First column: Low-light images, second column: heat map of initial illumination map, third column: heat map of estimated illumination map, fourth column: enhanced results, fifth column: denoised results via bilateral filtering.

Some low-light images from ExDark dataset, and results obtained from our implementation.

Citation

If you find this code helpful and use it in your research, please cite the following work:

@ARTICLE{guo_lime,
  author={X. {Guo} and Y. {Li} and H. {Ling}},
  journal={IEEE Transactions on Image Processing}, 
  title={LIME: Low-Light Image Enhancement via Illumination Map Estimation}, 
  year={2017},
  volume={26},
  number={2},
  pages={982-993},}