Active Contour Research

Course research on the active contour technique in computer vision.

Transcript

1. Active Contour Research
   by Nicolas McCurdy
   

   View Slide

2. What is active contour?
   ● A way of finding the curved outline of an
   object while avoiding any background noise.
   ● It works for 2D images, but similar
   techniques can be applied to 3D images.
   ● Also known as snakes.
   

   View Slide

3. How it works
   1. Create the snake (a connected series of
   vertices) at some distance from the object.
   2. Minimize the energy iteratively, which
   gradually moves the points of the snake
   closer to the contour of the object.
   3. Stop when the energy is minimized (the
   snake’s contour matches the object’s
   contour).
   

   View Slide

4. Visual example
   Source: http://commons.wikimedia.org/wiki/File:Snake-contour-example.jpg
   

   View Slide

5. Active contour with edges
   Edge detection algorithms are used on each
   iteration to determine where the object is (and
   to adjust the current contour appropriately).
   Advantages
   ● Existing edge detection algorithms can be
   used.
   

   View Slide

6. Active contour with edges
   Disadvantages
   ● Only objects with gradient edges can be
   detected.
   ● Usually, the points on the contour can only
   move in, not out.
   

   View Slide

7. Summary of “Active Contours Without Edges”
   This model is not based on a stopping edge
   detection function. Instead, the image is
   segmented by minimizing an energy.
   Advantages
   ● Works for images with and without gradients.
   ● Supports finding smooth edges and even
   disconnected edges.
   

   View Slide

8. Part 1: Create a model with a fitting term
   To find the curve, minimize the fitting term by
   moving the curve C closer to the boundary of
   the image in multiple iterations.
   Summary of “Active Contours Without Edges”
   

   View Slide

9. Summary of “Active Contours Without Edges”
   

   View Slide

10. Part 2: Mumford-Shah functional
    For this approach to active contour, the
    Mumford-Shah functional is used to help
    segment the image into multiple objects before
    the fitting term is minimized.
    Summary of “Active Contours Without Edges”
    

    View Slide

11. Part 3: Level set formulation
    The level set method can be used to solve the
    specific case of the minimal partition problem
    that results from segmenting the image with
    this technique. The technique described in this
    paper uses the level set method to translate the
    Mumford-Shah model to segment the image.
    Summary of “Active Contours Without Edges”
    

    View Slide

12. Conclusion
    ● The active contour’s points are computed
    and moved in iterations until the solution is
    stationary.
    ● This is the default method used in MATLAB’
    s activecontour function.
    Summary of “Active Contours Without Edges”
    

    View Slide

13. Example
    Summary of “Active Contours Without Edges”
    

    View Slide

14. Using activecontour in MATLAB
    activecontour(A,mask,n,method)
    ● A is a 2D grayscale image.
    ● mask is a binary image.
    ● n is the number of iterations.
    ● method is the method used for active
    contour (‘Chan-Vese’ or ‘edge’).
    

    View Slide

15. Using activecontour in MATLAB
    activecontour(A,mask)
    ● Defaults to 100 iterations.
    ● Defaults to the ‘Chan-Vese’ method (as
    described previously).
    

    View Slide

16. Using activecontour in MATLAB
    Example
    I = imread('coins.png');
    mask = zeros(size(I));
    mask(25:end-25,25:end-25) = 1;
    bw = activecontour(I,mask,300);
    figure, imshow(bw);
    title('Segmented Image');
    

    View Slide

17. Using activecontour in MATLAB
    Example Results
    

    View Slide

18. References
    ● MATLAB’s activecontour documentation
    ● Papers
    ○ [1] T. F. Chan, L. A. Vese, Active contours without
    edges. IEEE Transactions on Image Processing,
    Volume 10, Issue 2, pp. 266--‐277, 2001
    ○ [2] V. Caselles, R. Kimmel, G. Sapiro, Geodesic
    active contours. International Journal of Computer
    Vision, Volume 22, Issue 1, pp. 61--‐79, 1997.
    

    View Slide