Amoeba

An Image Segmentation Algoritm

Amoeba Home
Screen Shots
Download
Project Report
Links

Amoeba is an active contour based image segmentation alogritm proposed by Iannizzotto, G and Vita, L [1].

Image segmentation can be broadly defined as partitioning an image into its constituent objects or parts. Amoeba is an active contour based image segmentation algoritm. Active contours are closed chains that have a high degree of morphological adaptation.

This implementation runs on Linux and requires Gnome libraries to be present. You need not be running Gnome for Amoeba to work. The project has been designed using Glade and the source code has been taken directly from the the project directory.
The basic backend code is written in pure ANSI C, and should run on any OS. Only you need to detach it from the rest of the code. I had the code but it was lost with a disk crash. The current code is all that was salvaged from the backup.

The significant features of this implementation include:

  • O(n): Amoeba is one of the fastest algoritm image segmentation algoritms.
  • Full Graphical interface
  • All important parameters can be varied
  • Images can be loaded from many formats.
  • Choice of Various Gradient Operators

BUGS
This source code has been salvaged from a disk crash and is about ten days older than the final version. Some of the last minute features are missing, but they are not critical to the functioning of the project. I have no immediate or future plans to reimplement the missing features or kill the bugs. A list of missing features and bugs is here.

  • Documentation: The attached documentation is in postscript format. Once there existed an Amoeba.lyx file from which Amoeba.ps was generated. Also some pages are missing from the ps file.
  • Laplacian Operator: This operator is not implemented completely in this release.
  • Recognition: Image recognition is implemented for very simple geometrical objects only. There were some more shapes implemented at the last minute. Do not run the Recognition for complex images.

  1. Fast and Accurate Edge Based Segmentation with no contour smoothing in 2D Real Images, IEEE Trans. Image Processing, vol 9, pp. 1232-1237, July 2000

This project was developed by Me(Ajay kumar Dwivedi) and Harikesh Singh under the able guidence of Dr. Govind Sharma , Associate Professor, IIT kanpur. This project won the best "B. Tech. Project in Electrical Engineering" for the year 2001

SourceForge Logo