The Intelligent Scissors can be used to select an image region defined by strong color-changes at the edges.
You can learn more about it in papers:
pip install intelligent-scissors
To use in your program
from scissors.feature_extraction import Scissors
image = ...
scissors = Scissors(image)
seed_x, seed_y = ...
free_x, free_y = ...
path = scissors.find_path(seed_x, seed_y, free_x, free_y)Also you can run a simple demo
from scissors.gui import run_demo
file_name = 'image.png'
run_demo(file_name)The current implementation includes
- Cursor snap
- Static features
- Dynamic features
- On-the-fly Training
- Unrestricted graph search
On-the-fly Training allows you to select a “good” initial-boundary segment. However, this results in poor performance for new segments with different intensity/gradient magnitudes. To overcome this, you should first try to select a small region of the new segment to create correct dynamic features.