This software package is
provided for research purpose only. Please cite the paper below for
Zhuowen Tu, "An Integrated Framework for Image Segmentation and Perceputal Grouping", ICCV, Oct, Beijing, 2005.
Note: This is a faster version than that reported in
Zhuowen Tu and Song-Chun Zhu, "Image Segmentation by Data-Driven Markov Chain Monte Carlo", IEEE PAMI, Vol. 24, No. 5, May, 2002.
But the two produce similar results as they share the same underlying models.
This package is for Windows ONLY and very easy to use and uses the FreeImage library for reading images of general format. It currently does not include color models though you still can use it to segment your color images.
Usage: [input image file] [output path] [-t number of iterations] [-s scale (default=3.0)] [-r 0:random seed 1: fixed seed] \n
input image file: either .bmp or uncompressed .tif file
output path: the path to save the results
-s scale: the scale factor
-r: Since this is an MCMC approach, you may not get indentical results each time. In order to get the same ones under certain situations, you may set this value to 1. The default value is 0.
There are three files saved to the path you specify:
(1) your_name_boundary.tif : boundaries for the segmented regions
(2) your_name_label.tif: label map for the segmentation
(3) your_name_sublabel.tif: the label map for the atomic regions as described in the paper.