

					-----------------

					PATTERN Directory

					-----------------



File Descriptions

*****************



The PATTERN directory documents codes that are used to perform image pattern 

recognition. The first (and most prominent) set files in the PATTERN 

directory pertains to the k-medoid-model, and is labeled as such.



Aside from this set of file, there are two model descriptions that are 

included in the PATTERN image-processing directory. The first is the IMGKMED 

model, based on the k-medoid model. It is described in the PDF document with 

the same name. Included in the description is a generic MATLAB code.



The second is a text input file for running the Benabdallah-and-Wright 

pattern-recognition model. Depending on the mathematical-programming package, 

this file can either be directly executed by the optimization software, or it 

can be executed after minor editing in format.





Operating Instructions for the k-Medoid Model

*********************************************



o	Open the MATLAB software

o	Go to the "MATLAB-k-medoid-code" directory/folder in which you have 

placed the "imgkmed" software. At the command line, enter imgkmed (in lower 

case)

o	Press the "Load image" button

o	Open the "testg.bmp" file

o	Set k = 3

o	You can set the scale for `proximity' or for Channel 1 (monochromatic 

grayscale) by moving the lever. Since we are operating on monochromatic 

grayscale, simply leave the Channel-1 scale at the middle. When setting the 

weights, remember the right end of the `proximity' scale suggests contextual 

classification.

o	Determine the number of iterations required to reach algorithmic 

convergence.

o	Execute by pressing the `Cluster' button.

o	Reload image for another run.





Notice that since the provided image(s) is black-and-white, channels 2 and 3 

are non-functional. For simplicity, please leave the channels 1, 2 and 3 

settings to the middle point (50-50). I also suggest that you leave the 

setting for `Proximity' at the half way point when you start out. Later on, 

when you decide to change w1 and w2 (at the `proximity' setting). Remember w1 

is the weight on gray value, and w2 is the weight on proximity.



When you run the IMGKMED software, make sure you re-load the original "test" 

image every time. In other words, anytime you change your input parameters, 

such as the number of iterations, you need to re-load the "test" image. 

Otherwise, erratic behavior will result from the IMGKMED software.



					*	*	*



The IMGKMED software was developed by Terra D. Colvin in 2007, and modified 

in 2009 by Jerzy S Zielinski. The software was supervised by Yupo Chan at the 

University of Arkansas at Little Rock.

