This package contains an implementation of Ljosa and Singh's probabilistic segmentation algorithm, described in the following paper:
Vebjorn Ljosa and Ambuj K. Singh: "Probabilistic Segmentation and Analysis of Horizontal Cells," in Proceedings of the 2006 IEEE International Conference on Data Mining (ICDM), December, 2006, doi:10.1109/ICDM.2006.129.
The program is written in C, and simulates a random walk on the image. We also have an eigenvector-based solution, but that is not included in this package at this time. We may incorporate it in the future. A Matlab MEX interface is provided.
The software is released under the BSD license. See the file LICENSE for details.
The latest version is 1.0, and can be downloaded here: http://www.ljosa.com/~ljosa/software/ljosaprobseg/download/ljosaprobseg-1.0.tar.gz.
The individual source code files are available here, in case you would like to take a look at the code before downloading.
Just typing make should build the stand-alone program ljosaprobsegsim as well as a Matlab MEX module called ljosaprobseg. The stand-along program can, for instance, be invoked as follows:
./ljosaprobsegsim -o output.pgm -n 2 -8 -r 0.0001 7cap33x.pgm \
5000 206,467 207,467 206,468
Do ./ljosaprobsegsim -h to learn about the available options. Type help ljosaprobseg in Matlab to learn how to invoke the Matlab module, or see the example in demo_ljosaprobseg.m. The image used in the demo was provided by Geoffrey P. Lewis from the laboratory of Steven K. Fisher at UCSB.
The programs are multithreaded in order to improve their running time on machines with multiple processors or cores. For the stand-alone program, use the "-n" option to specify the number of worker threads. The Matlab module is hardcoded to use two worker threads. (This can easily be changed in the source code, if desired.)
Contact Vebjorn Ljosa with any questions.