This repository is the implementation of the paper Flow-Based Image Abstraction by Henry Kang, Seungyong Lee and Charles K. Chui.

python3 run.py --file FILE_PATH

By default the program will generate 5 images with different amount of edges. The following flags can be used.

--threading to enable threading. This will create separate threads for FBL and FDoG. Also FDoG for different values of r will also be calculated in separate threads. Though we didn't see much speedup by enabling threading.

--greyscale flag will remove chrominance from the image resulting in greyscale image.

The program takes around 5 min for 600*800 images. If you have a very large image, please reduce the size.

Other flags will be added later

Here are few of the samples generated so far. The leftmost is the original image, the middle one and the right figures are generated using r = 0.5 and 0.8 for FDoG.

The left image is the original whereas the right image is the best looking among r = 0.5,0.6,0.7,0.8

The following were created by running the above program on each frame of video and then merging them.