Procedurally generated pirate treasure maps. X marks the spot!

I used several excellent third party libraries...

• cairo for rendering
• colour for color interpolation
• noise for simplex noise
• Pillow for saving debug images of noise layers
• pyhull for delaunay triangulation
• Shapely for all kinds of 2D geometry operations

The script will generate several random maps and save them as PNG files.

git clone https://github.com/fogleman/PirateMap.git
cd PirateMap
pip install -r requirements.txt
python main.py

It took me a while to decide on an approach for generating the land masses. I didn't want to just generate some simplex noise, threshold it and render it. I wanted to actually compute a polygonal shape that I could do further operations on. So here's how that works...

• generate a layer of simplex noise with several octaves, with decreasing values as the edge of the image is approached
• fill the screen with randomly positioned points evenly spaced using the poisson disc algorithm
• filter the poisson points to those where the corresponding noise value is above some threshold
• take these points and compute a "concave" hull or alpha shape with them

Shapely's buffer function is used heavily for padding or cleaning up shapes.

I also wrote an xkcdify function to add some perturbations to some of the polygons, namely the different water color gradations.