Automatic Color Correction using Colour-Checker Detection

Overview

This repository implements an end-to-end color correction pipeline based on automatic ColorChecker detection.
It uses the open-source colour-science and colour-checker-detection libraries to detect the ColorChecker and apply color correction to the image.

The workflow covers every stage of the color correction process: ColorChecker detection, patch sampling, RGB linearization, color conversion, correction matrix derivation, and output verification.

Objective

The aim of this exercise is to reproduce a practical camera color correction workflow in Python.

It shows how to:

• Detect a ColorChecker automatically from a scene image.
• Extract patch values and compute average RGB data of each patch.
• Convert nonlinear sRGB values to linear RGB values.
• Compute both linear (3×3) and second-order polynomial (3×9) color correction matrices.
• Apply and evaluate the resulting correction matrices.

Getting Started

1. Clone this repository

git clone https://github.com/<kimiarfaie>/Color-Correction.git

2. Install the required packages

Make sure you have Python 3.12 or later installed. All dependencies are listed in requirements.txt. Install them with:

pip install -r requirements.txt

3. Run the Jupiter Notebook

Place your input images inside the Images/ folder and set the correct path inside the notebook.

Then open and run the notebook step by step.

Output Files

After running the notebook or script, the following files will be generated inside the output directory:

• rgb_mean_norm_colors.csv — normalized RGB patch means
• rgb_mean_norm_linear_colors.csv — linearized RGB patch means
• XYZ_reference_ccaft2014_D65.csv — reference XYZ values (D65 adapted)
• Lab_reference_ccaft2014_D65.csv — reference Lab values (D65 adapted)
• CorrMatrix_Linear.csv — 3×3 linear correction matrix
• CorrMatrix_Poly2nd.csv — 3×9 second-order polynomial correction matrix