Denoiser Comparator

Little helper framework to compare image denoisers. The structure is split in components that can be selected via command line arguments. The following component types are available:

• denoiser: Each denoising technic is implemented as an individual denoiser that can be selected at runtime
• metric: Noise reduction can be measured through various metrics (some priorize the color discrepancy, others take into account human eye perception, etc). A few different metrics can be enabled.
• dataset: Multiple image datasets can be implemented, but only one can be used at a time

Structure

• denoisers/ has the denoiser implementations
• metrics/ has the metrics implemented
• datasets/ has the datasets implemented

Requirements

The best way to make sure you are running with all the required dependencies is to create a venv and install all the requirements there:

# create the venv
python3 -m venv name_of_venv

# activate the venv
source name_of_venv/bin/activate

# install the requirements
pip install -r requirements.txt

Note on BM3D

Apart from the python modules installed via pip, if you want to use the bm3d denoiser, you need to install the OpenBLAS development package (openblas-devel on OpenSUSE and Fedora, libopenblas-dev on Debian and Ubuntu).

Running

You can check the available command line arguments by calling:

./denoise_comparator.py --help

If you would like to see the available components, just invoke the main script with --list:

./denoise_comparator.py --list

Enjoy!