10000 pixels partitions in to 30 regions as follows:
415, 91, 596, 420, 347, 239, 313, 298, 447, 259,
414, 229, 137, 347, 353, 449, 744, 453, 479, 313,
362, 363, 494, 87, 343, 395, 327, 140, 62, 285
Right now, this uses LLoyd's algorithm or k-means clustering on N=104 points, where we have decided the centers in advance.
- In an earlier version, the mapping of the points to the clusters was not parallel. Fixed
- Naive implementations tend to run in N2 time while Fortune's Sweep Line runs in N log N.
- There may be other time saving tricks as well:
scicomp.stackexchange.com
.
We can apply this to get Picasso-like images from normal photographs.
Amazing! You are an artist.