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.

If you're using MacOS:

Install LLVM with:
```
brew install llvm
```
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
```
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.
To run tests for C- extensions, execute
```
pytest test
```
under project folder.

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