Python-SRM — Statistical Region Merging in Python

Description

Python-SRM is a Python implementation of the Statistical Region Merging algorithm (SRM), as described in TPAMI'04 or PR'05 (see Bibliography for more details).

See the Java applet for an online demonstration of the method.

Prerequisite

numpy

Usage

The SRM module contains one class called SRM. The class constructor expects two arguments:

a 3-channels array containing the image data,
a rough estimate of the number of expected regions.

After the initialization, the segmentation can be launched with the method run which expects no arguments.

See the following section for an example.

Example

from scipy.misc import imread
from matplotlib import pyplot

from SRM import SRM

im = imread("lena.png")

srm = SRM(im, 256)
segmented = srm.run()

pyplot.imshow(segmented/256)
pyplot.show()

Download

Directly: http://www-connex.lip6.fr/~schwander/python-srm/python-srm-0.1.tar.gz

Darcs repository: http://www-connex.lip6.fr/~schwander/darcs/python-srm/

Bibliography

1. Nock, F. Nielsen: Semi-supervised statistical region refinement for color image segmentation. Pattern Recognition 38(6): 835-846 (2005)
1. Nock, F. Nielsen: Statistical Region Merging. IEEE Trans. Pattern Anal. Mach. Intell. 26(11): 1452-1458 (2004)
1. Nielsen: Visual Computing: Geometry, Graphics, and Vision. Charles River Media, ISBN 1-58450-427-7 (2005)

Other implementations

A Java one by one of the original authors, F. Nielsen: http://www.lix.polytechnique.fr/~nielsen/Srmjava.java
A Matlab implemention by S. Boltz: http://www.mathworks.com/matlabcentral/fileexchange/25619-image-segmentation-using-statistical-region-merging

Licence

Python-srm is free software, released under the term of the BSD license.