This project implements (currently not, hopefully soon) the impulse noise (also known as salt-and-papper noise) removal algorithm published in paper entitled A weighted mean filter with spatial-bias elimination for impulse noise removal (UWMF, see here). UWMF is based on another paper entitled Interpolation-based impulse noise removal (IBINR, see here). IBINR basically interpolates the corrupted pixel using inverse Euclidean Distance. UWMF recalculates the weights (based on inverse Manhattan Distance) according to the spatial distribution of corrupted pixels, effectively yielding a more spatially-balanced weight contribution.
This will most likely be a command line tool. I am planning to make a CUDA implementation once this one is finished.
- test new option parser
- make sure to use release builds of zlib and libpng
GPLv3. See LICENSE for full description.