Skip to content

Latest commit

 

History

History
34 lines (23 loc) · 764 Bytes

README.md

File metadata and controls

34 lines (23 loc) · 764 Bytes

Non Local Means - CUDA

Implementation of Non Local Means algorithm for image denoising.

  • version 1: CPU sequential in C++
  • version 2: GPU parallel in CUDA
  • version 3: GPU parallel using shared memory in CUDA

Build instructions

Requierements: nvcc (Nvidia CUDA Compiler)

Set parameters in Makefile and

make

or

mkdir -p data/out
nvcc -o build/main -I./include src/main.cu -O3

and run with

./build/main $(imageNum) $(patchSize) $(filterSigma) $(patchSigma) $(useGpu) $(useSharedMem)

Instructions to see the filtered images

Run the script matlab/show_image.m in MATLAB and all the output images of folder data/out will show up.

Original and noisy images can be found in data/images.