Quick Alignment for Burst Photography

This repository contains a multi-threading implementation of L2/L1 mixed alignment algorithm described in Hasinoff et. al, HDR+, SIGGRAPH'16.

Author: Rundong Li <lird@shanghaitech.edu.cn>

Get Started

Note this implementation gets NO WARRANTY of running on Windows platform. MacOS with LLVM >= 6.0 or Ubuntu with GCC >= 4.9 is required.

0. If you're using MacOS:

   Install LLVM with:

   brew install llvm

1. Install required Python extensions by pip or anaconda, note that Python >= 3.6 is required.

   conda install -r requirements.txt

   Then you may build C- extensions by

   ./build.sh

2. Create a folder data and copy your burst into it, execute

   python faster_align_cpu.py

   then you should find the results plotted in data/demo.png.

   Note: the demo.py is just for illustrating the algorithms we used, itself can work too, but very slow.

3. To run tests for C- extensions, execute

   pytest test

   under project folder.

License

This work is licensed under GNU GPLv3, see LICENSE for details.