A little tool of python to divide graphs into regions through clustering.
To use it you need to download the whole package but the functioning part can be found in the file regionDividor.py
Package requirement: multiprocessing; cv2; threading; numpy; PIL if image type convertion function is used;
Performance: linear time to the number of points in the image; employed multiprocessing; performance requires more improvement: i.e.: the function needs around 30 seconds to process a 353*546 image of approximately 40 clusters (US map is used to conduct the test on 4-core i7 processor Mac pro 15 inch); a crude way of image compression provided in utils, only support compression factor of integers and the data size will shrink at least 4 times compared to the original one (the portion of the data points left is in the order of compression factor squared) a compression method based on portion provided in utils.
Possible update plans:
faster algorithm of cluster diffusion from central point, previous
attemp on multiprocessing is failed due to failure to overcome
the performance loss of process initiating and communication;
re-write some key computationally intense tasks in faster languages
like C, C++;
finding a solution to the technical flaw that the algorithm does not
recognize dotted lines as boundary, image may require additional
prior processing;
Little Note: the function finding the gradient is named as fast because it has been improved with multiprocessing