9/17/97

File "segment" is the binary executable for color segment algorithm.
It was compiled on SGI IRIX 5.3 operating system.
The algorithm is developed by Wei-Ying Ma and B. S. Manjunath 
at University of California, Santa Barbara.
For more details, please read the following web document:
http://vivaldi.ece.ucsb.edu/Netra/

To run the program, type "segment" to show a list of command line arguments. 

The <5> argument "offset" is the parameter to determine
the scale factor of segmentation. For image size 256x256, offset
6 - 8 is suggested. For image size 128x128, use value around 4.
The best value of offset depends on the image. Please try around
to find the optimal value for the images you are using.

The <6> argument "estimated number of regions" is not very effective.
Give the minimum value you prefer. The program will segment the image 
into minimum number of regions under the specified "offset" value, 
which is usually higher than the estimated number of regions you specified.

This program takes quite amount of memory. Please use your high-end
workstation for processing.

The output the program is a file with extension "edge". It is the
edge map of segmented image. It is in compressed format. In order
to view the result. Use program "getregion" provided. The output
will be two files. One is the region label of the original image
with extension "rmap". It is a raw gray-scale image and can be viewed.  
The other file is the edge superimposed on the original image with extension
"edg2". It is in raw rgb format. Type "getregion" to see 
command line arguments.

There are free image viewing utilities to view raw image data.
For example, "display" from ImageMagick. You can search it 
through web. If you use "display" type the following command:
display -size widxhei rgb:filename

File "sample.rgb" is a small sample color image for your testing. 
It is 192x128 (wid x hei) in rgb raw format.

Example to run:

> segment 1 sample.rgb 128 192 8 4 . .

Output: sample.rgb.edge(edgemap)

> getregion sample.rgb 128 192 1

Output: sample.rgb.rmap(regionmap) sample.rgb.edg2(image with edge)

> display -size 192x128 rgb:sample.rgb.edg2