Active contour model (aka snakes) implemented in python.
An active contour model is an image analysis technique which evolves a curve to find boundaries in images. Skimage has an implementation of this algorithm, but like many of their functions, it does a lot of preprocessing "under the hood."
This code requires only the standard python scientific stack: numpy, scipy, matplotlib, and skimage. The main functions are fit_snake (and snake_energy) in BKlib.py; this function will evolve a closed contour to fit an image. Starting with an initial guess for the boundary points, fit_snake shifts the points around until they sit at a local minimum of the function snake_energy. The snake energy is pretty much as described in the wikipedia article: internal energy terms penalizing stretching, and encouraging smoothness; and external energy terms that push the curve towards edges.
I've included some examples code and data to demonstrate how it works. Just clone, and run python snakes_demo.py. It will show the results of the image preprocessing, initialize the snake/contour from one the of the supplied .npy data files, and show an animation of the fitting procedure.
These functions form the basis of the scripts in one of my other repos.