NetLogo banner

 Home
 Download
 Help
 Resources
 Extensions
 FAQ
 References
 Contact Us
 Donate

 Models:
 Library
 Community
 Modeling Commons

 User Manuals:
 Web
 Printable
 Chinese
 Czech
 Japanese

  Donate

NetLogo User Community Models

(back to the NetLogo User Community Models)

MASforSeg

by KHOUADJIA Mustapha Redouane (Submitted: 08/10/2007)

[screen shot]

Download MASforSeg
If clicking does not initiate a download, try right clicking or control clicking and choosing "Save" or "Download".(The run link is disabled because this model uses external files.)

A MULTI-AGENTS SYSTEM FOR THE IMAGES SEGMENTATION

WHAT IS IT ?

This program presents an adaptative approach based on a multi-agents system for the images segmentation. This system is constituted at the micro level of autonomous entities, which deployed on the image. They are equipped with a capacity to estimate the homogeneity of a region from their current locality. Each entity exhibits several behaviours in response to the local stimulus. It can migrate, reproduce, or diffuse within the image. Various entities explore the image and label the pixels when they belong to homogeneous segments. The interactions at the micro level allow the emergence of a new feature that is segmentation of the image.

HOW TO USE IT

1. Import an image to the PGM format(Image in level of grey) and fit the screen to the dimensions of this one
2. Fix the number of segmentation agents for every class (4 classes by default)
3. Adjust the perception radius of the agents R_region, Their life span, As well as the number of offspring which they can generate during the self reproduction
4. Adjust the step of migration. This step correspond to the dispersal amplitude of the segmentation agents
5. Create and distribute the segmentation agents on the image with the button "Distribution"
6.Throw the program with the button “go”

THING TO NOTICE

Each entity is equipped with an ability to estimate the homogeneity of a region from its locality. When an inhomogeneous segment is met, the autonomous entity diffuses itself towards a nearby pixel by taking a direction which it will have to select.
In case the entity locates a homogeneous segment, this last one, through the self reproduction behaviour, delivers a certain number of offspring in its local region, and labels the pixel on which it resides.
The new offspring will have the task to label the same type of segment as the parent entity.
In its birth, the offspring presents a behaviour of migration, if ever there is a strong concentration of congeners in its local region. A migration can also arise, in case, the pixel on which is the entity is already labelled.
Each entity has a predefined life span. When an agent takes age, it gains in maturity. This maturity is indicated by its current state (young, adult, parent, old).
It will help it to improve and acquire during its life of the knowledge, allowing it to return its search more effective.
The directions taken by the entity during its diffusion are determined by a vector of directions. The elements of this vector correspond to the probability of success to find a homogeneous segment; if a given direction is selected.
The probabilities are updated by basing itself on the directions taken beforehand by the better adapted entities belonging to the family of the entity.

THING TO TRY

Try to practice various steps of migration using slide. It will have for consequence of change the density of the entities swarm through the picture.

EXTENDING THE MODEL

Our system remains opened and several points can be studied:

-The possibility of automating some mechanisms of agents segmentation, such as the life span, and the number of generated offspring. The agent can so manage its parameters according to the local constraints to which it is confronted. For example: an agent will limit the number of created offspring, if ever there is a strong concentration of congener in its local region.

-Try to specialize our system for the detection of certain medical abnormalities at the level of Brain MRI. It will be question for example of defining segmentation agents whose role is the detection of tumours at the first stage.

CREDITS AND REFERENCES

Author: KHOUADJIA Mustapha Redouane
Affiliation: Faculty of Engineering, Computer Science Institute, Constatine University, Algeria.

Email:khouadjia@yhaoo.fr

To refer to this model in academic publications, please use:

M.Khouadjia, S.Meshoul, E-G.Talbi: "A Complex System for the Images Segmentation" in META06, workshop on metaheuristics, Hammamet November 2-4 2006, Tunisia.

NOTES

You can create easily images in the PGM format by means of the Xnview software :
http://www.xnview.com/

(back to the NetLogo User Community Models)