CS 7322: Computer Vision II Spring 1997 (aka. High-Level Computer Vision) Mid Term Project Handed Out: 4/21/97 Due Back: 5/12/97 (3pm) Warning: NO EXTENSIONS

Snakes: Active Contour Models

Implement snakes. Here are some pointers that would help.

1. Use class notes and/or the original Snakes paper (Kass et.al.).
2. There is a snakes demo available from Matlab (www.mathworks.com/contsoft.shtml, search for "snakes") OR from class www pages (under midterm/snakes). Try this one out and see what you understand. What kind of an algorithm is used here ?
3. Look around and see what other alogorithms exist. Remember the WWW site I should you for keyword searches: (www.cs.cmu.edu/afs/cs/project/cil/ftp/html/v-pubs.html). Implement an alogoritm based on your favorite one. Just try looking around on the WWW there is a lot of stuff out there.
4. What I am looking for is an implementation beyond the matlab demos in point 1 above. You can use the graphics module in that demo tho. Your program should be much different.

Here are the tests I'd like you to do after you have finished your implentation (ie. keep them in moid for implementation).

1. Test your program to find contours in four monochrome images of your choice. To make testing easier, one of your test images should probably be an artificial image (e.g., a white disk or square on a black background). Be sure to save "before" and "after" shots that show the snake in its initial and rest position.
2. Test your program on some real images. Try out various pre-processing steps and see what helps.
3. Use your snake module to track a moving object over a number of frames in an image sequence. The user should only need to place a "template" on the first image. All subsequent images in the sequence should use the previous frame's snake as an initial model.

What to hand in (on 5/12/97. by 3pm, via the WWW)

1. Create an HTML document that gives the matlab program source and output for your example cases.
2. Run your snakes module and the demo version over a series of image/image sequences. Evaluate the results. Include both the before and after shots.
3. What algorithm did the snake demo use? Explain it? What are its limitations
4. What does your method do that is different then the give one? Explain it? What are the limitations?
5. How sensitive are the algorithms to setting the parameters alpha and beta? Why?
6. How stable are the algorithms with respect to initial placement of the snake in the image? Why?
7. How stable are the algorithm with respect to the time step? Why?

There is a template in your WWW dir for this class under midterm/report.html. An example is available from class WWW pages midterm/report.html.

Additional Info.

• Class notes. (Irfan Essa, Robert Sumner) Also see ZenPad
• Examples of Working Systems using Snakes
  • Medical Application (U of Washington)
  • Active Snakes (UK)
  • Active Contours, Java demo
  • Multiscale Snakes (NYU)
  • Many more ... if you find 'em let me know.
• References:
  • M. Kass, A. Witkin and D. Terzopoulos, Snakes: Active contour models, Int. J. Computer Vision 1, 1988, 321-331
  • D. Williams and M. Shah, A fast algorithm for active contours and curvature estimation, Computer Vision, Graphics, and Image Processing: Image Understanding 55, 1992, 14-26
  • Amini, A.A., Tehrani, S., and Weymouth, T.E., Using Dynamic Programming for Minimizing the Energy of Active Contours in the Presence of Hard Constraints, ICCV88(95-99).
  • For more see: "http://iris.usc.edu/Vision-Notes/bibliography/segment181.html"
  • Just a "search" for snakes in USC vision bibliography suggests the following titles.
    • Active Contours, Snakes or Deformable Curves
    • Active Shape Models: Smart Snakes
    • Active Volumes, Deformable Solids, 3-D Snakes, etc.
    • B-Snakes: Implementation and Application to Stereo
    • Catching Moving-Objects with Snakes for Motion Tracking
    • Controlling Growing Snakes by Using Key-Points
    • Darboux Frames, Snakes, and Super-Quadrics: Geometry from the Bottom Up
    • Deformable B-Solids and Implicit Snakes for Localization and Tracking of MRI-SPAMM Data
    • From Ziplock Snakes to Velcro(TM) Surfaces
    • Gradient Vector Flow: A New External Force For Snakes
    • Imposing Hard Constraints on Soft Snakes
    • Initializing Snakes
    • Making Snakes Converge from Minimal Initialization
    • Medical Image Segmentation Using Topologically Adaptable Snakes
    • On Regularization, Formulation, and Initialization of the Active contour models (Snakes)
    • Region Competition: Unifying Snakes, Region Growing, and Bayes/MDL for Multi-band Image Segmentation
    • Region Competition: Unifying Snakes, Region Growing, and Bayes/MDL for Multiband Image Segmentation
    • Semi-Automatic Feature Extraction by Snakes
    • Semi-Automatic System for Edge Tracking with Snakes, A
    • Snakes and Splines for Tracking Non-Rigid Heart Motion
    • Snakes, Motion Tracking
    • Snakes, Restricted Curves, Splines, etc.
    • Snakes: Active Contour Models
    • Spherical Winged B-snakes
    • Statistical Snakes: Active Region Models
    • Topologically Adaptable Snakes
    • Vector (Self) Snakes: A Geometric framework for Color, Texture and Multiscale Image Segmentation
    • Ziplock Snakes

Solutions

• Causenge Cangin
• Winncy Y. Du
• Andrew B. Gardner
• Jose Gerardo Gonzalez
• Roman S. Khramets
• Roberto Peon
• Wasinee Rungsarityotin
• Timothy E. Smith
• Mark Snider
• Scott Stillman
• Robert Walker Sumner
• Lisa Wenner

Last Changed: 5/7/97 12:30pm by Irfan
File: /net/www/classes/cs7322_97_spring/midterm/midterm.html
WWW: www.cc.gatech.edu/classes/cs7322_97_spring/midterm/midterm.html