MVRegFus

Multi-View Registration and Fusion

Python module to process multi-view light sheet data. This includes

Image registration
- view registration
- time registration (e.g. for drift correction)
- channel registration (affine chromatic aberration correction)
View fusion
- weighted additive fusion
- multi-view deconvolution (Richardson-Lucy)
- for both fusion types, weights can be:
  - blending weights
  - based on an image quality metric

Additional features:
- processing pipeline is leveraged by dask
  - entire workflow can be executed on laptops, independently of output dataset size (two input views need to fit into memory at once)
  - execution on computational clusters using dask.distributed
- GPU accelerated multi-view deconvolution (using CuPy)

Notes:

MVRegFus uses the elastix registration toolkit for performing most image registration.

Installation instructions

clone this repository from github: git clone https://github.com/m-albert/MVRegFus.git
use anaconda to install an environment from the provided .yml file: conda env create --file MVRegFus/mv_environment.yml
install MVRegFus pip install -e ./MVRegFus
Download elastix 4.9 (binary version 4.9 suitable for your platform) and place files into a folder 'elastix' next to the project folder, as such:

./MVRegFus
./MVRegFus/bin
...
./elastix
./elastix/bin
./elastix/lib

You're set!

open terminal / anaconda prompt
activate the previously installed anaconda environment: conda activate mvregfus (Win,MacOS) or source activate mvregfus (Linux)
copy mvregfus/bin/mvregfus_bin.py to <your_mvregfus_bin.py> (can be placed next to data) and use as a template to indicate the location of your files, resolution, etc. (see comments in file). Soon configuration file handling will be added
run fusion from the terminal with python <your_mvregfus_bin.py>

Python libraries:

External: