knnimdenoiser

KNN image denoise is image denoising using K Nearest Neighbors filter.

Overview

This application uses a K Nearest Neighbors filter to remove noise from images. The project was based on this paper.

Compilation

You can compile the application using a command:

make

Running the application

In order to run the program, you need to specify two arguments and options:

usage:
./knnimdenoiser [options] in.png out.png

options:
  -c N.N    noise coefficient (default 3.000000)
  -l N.N    lerpc (default 0.160000)
  -n N.N    noise (default 0.000000)
  -p NUM    threads count (default 2)
  -r NUM    radius (default 3)
  -t N.N    threshold lerp (default 0.660000)
  -w N.N    threshold weight (default 0.020000)
  -h        show this help message and exit

For example:

./knnimdenoiser input.png output.png

Example images before and after denoising process:

See knnimdenoiser-demo:

Noisy image	Fixed image