Question about light map details.

[bo-oseng]

First of all, thank you for your amazing research and for sharing your code.

Since I'm new to working with low light data, I might be asking something obvious. I want to create low light synthetic data from my data. I noticed there are parameters (mu, rho, ds, ss) that need to be set when obtaining the light map using "LIME: Low-Light Image Enhancement via Illumination Map Estimation".

I would like to know how you determined these parameters specifically when creating the CDD dataset. Could you please provide detailed information about the light map and LIME?

[wann-li]

I have the same problem

[gy65896]

This link explains how to generate the refined illumination maps (Tout) https://github.com/estija/LIME/tree/master.

[bo-oseng]

Thank you for your reply!

As I understand, when executing LIME's lime_main_module, parameters such as mu, rho, ds, and ss are required, and LIME provides several optimized hyperparameters. I have tried using the parameters for 'building' (mu=0.01, rho=1.188, ds=10, ss=1.5), but I'm not confident in their suitability. Could you advise on which set of parameters would best match the light maps in the CDD-11 dataset?

Would it be appropriate to use lime_loop.m or lime_bf_loop.m to determine suitable parameters for each image?

Code I have tried

addpath('./BM3D');
addpath('./CDD-11_test/clear');
img_in = imread('00002.png'); 

 #  LIME's optimized hyperparameters 'building' (mu=0.01, rho=1.188, ds=10, ss=1.5)
[Ti, Tout, img_out, Iout] = lime_main_module(img_in, mu=0.01, rho=1.188, ds=10, ss=1.5, flag=0);