baidut / BIMEF

Code and data for the research paper "A Bio-Inspired Multi-Exposure Fusion Framework for Low-light Image Enhancement" (Submitted to IEEE Transactions on Cybernetics)

BIMEF

Code for our paper "A Bio-Inspired Multi-Exposure Fusion Framework for Low-light Image Enhancement"

• The code for the comparison method is also provided, see lowlight
• Downloads: google Drive (Just unzip data to current folder)
  • Datasets VV, LIME, NPE, NPE-ex1, NPE-ex2, NPE-ex3, DICM, MEF
  • Since some methods are quite time-consuming, we also provide their results (e.g. [e-mail])
  • Since some metrics are quite time-consuming, we also provide their results (TestReport.zip)
• All the experiments can be reproduced easily by running experiments.m

From left to right: input images, results of MSRCR, Dong, NPE, LIME, MF, SRIE, and BIMEF(ours).

Datasets

• VV （**Busting image enhancement and tone-mapping algorithms: **A collection of the most challenging cases）
• LIME-data
• NPE-data, NPE-ex1, NPE-ex2, NPE-ex3
• DICM —— 69 captured images from commercial digital cameras: Download (15.3 MB)
• MEF dataset

Prerequisites

• Original code is tested on Matlab 2016b 64bit, Windows 10.
• matlabPyrTools is required to run VIF metric (vif.m).

Setup

Run startup.m to add required path, then you are able to try the following demo.

I = imread('yellowlily.jpg');
J = BIMEF(I); 
subplot 121; imshow(I); title('Original Image');
subplot 122; imshow(J); title('Enhanced Result');

Replace BIMEF with other methods you want to test.

Directory Structure

.
├── data           # put your datasets here
│   ├── MEF        # dataset name (VV, LIME, NPE...)
│        ├── out   
│        │    ├── loe100x100           # LOE visualization results
│        │    ├── TestReport.csv       # results of metrics
│        │    ├── TestReport__xxxx.csv # backups of TestReport
│        │    └── xxx__method.PNG      # output images
│        └── xxx.jpg                   # input images
│
├── lowlight       # lowlight image enhancement methods
├── quality        # image quality metrics (blind or full-reference)
├── util           # provide commonly used utility functions
│
├── demo.m         # simple demo of lowlight enhancement
├── experiments.m  # reproduce our experiments
└── startup.m      # for installation

Usage

Run experiments.

% specify datasets
dataset = {'VV' 'LIME' 'NPE' 'NPE-ex1' 'NPE-ex2' 'NPE-ex3' 'MEF' 'DICM'};
dataset = strcat('data', filesep, dataset, filesep, '*.*');

% specify methods and metrics
method = {@multiscaleRetinex @dong @npe @lime @mf @srie @BIMEF};
metric = {@loe100x100 @vif};

for d = dataset, data = d{1};
    data,  
    Test = TestImage(data);        
    Test.Method = method; 
    Test.Metric = metric;
    
    % run test and display results
    Test,                     
    
    % save test to a .csv file
    save(Test);
end

Show test reports.

% table
TestReport('TestReport__VV.csv'),

% boxplot
TestReport('TestReport__MEF.csv').boxplot;

Our method (BIMEF) has the lowest LOE and the highest VIF.