Blind Chromatic Aberration Correction with False Color Filtering


Aberrated image	Filtered result

This is a repository containing a non-official Cython-based implementation of the IEEE Transactions on Image Processing 2013 article Correction of Axial and Lateral Chromatic Aberration With False Color Filter by Joonyoung Chang, Hee Kang and Moon Gi Kang.

The method consists in a 1D filter independently run on the columns and rows of an image containing chromatic aberrations. Merging the horizontally and vertically filtered outputs yields the final restored image. Our implementation leverages the multi-threading abilities of Cython to achieve restoration on large images in less than 1 second.

This implementation is part of an IPOL paper describing in the detail the method. If this code is useful to your research please cite our paper [paper (to appear)][demo]

News

10/19: Try the online demo using this code!

Testing the code

You can test the code with a test JPEG image containing chromatic aberrations. First, compile Cython code and install the python package with

bash compile_cython.sh

This will create in the chromatic_aberration_filtering directory a filter_cython.c file. The file is already shipped with this code. If you do not want to run cython, simply set to False the USE_CYTHON argument in setup.py. Second, you can restore the image with the command line:

python main.py

This will save a 8-bit PNG file containing the restored image. The method can be also used on 16-bit TIFF images or straight after demosaicking/denoising in a typical ISP pipeline.

Requirements

Please run the pip install command below to install the requirements:

pip install requirements.txt

Description

The code contains two main modules: filter.py that is a numpy-pure implemention of the paper (and is extremely slow), and filter_cython.pyx that contains the sources for the Cython implementation (recommanded).

After installation/compilation, you can import in any project this code by importing the package

import chromatic_aberration_filtering

The actual function is

chromatic_aberration_filtering.chromatic_removal

Please refer to main.py to see an example of use.

Contact