ColorBlock: Parallel Image Style Transformation

This is a course project of CMU 15-618: Parallel Computer Architecture and Programming.

It uses CUDA accelerated K-means, OpenMP accelerated BFS, and CUDA accelerated edge smoothing to transform the image into color-block/glass-painting style. Its image processing engine is written in C++ without using any image processing library. Sequential codes are also provided in order to make validations and performance comparisons.

Authors

Dongsheng Yang, Xuyang Fang

Requirements

pillow
MPI
CUDA

Running

make
sh oilPainting.sh

You can change the input image and the parameters in the script.