10

I am working with MRI images of the brain that have certain areas marked by hand like this and this.

I am trying to come up with an interpolating function that will let me describe such curves so that I can use machine learning techniques to automatically generate such markings on unmarked images like this and this.

I've considered using cubic spline interpolations to approximate the curve, but as far as I know, I would have to split the curve into multiple parts to interpolate it.

Am I on the right track or are there other (better?) ways to do this?

rdasxy
  • 201
  • 1
  • 4
  • @rdaszy, you might want to check this Matlab example: http://www.mathworks.com/products/image/demos.html?file=/products/demos/shipping/images/ipexcell.html – Dov Mar 09 '12 at 03:42
  • There is a toolbox for image fitting: http://www.mathworks.es/products/curvefitting/ You can find some examples here: http://www.mathworks.es/support/tech-notes/1500/1508.html – Luis Andrés García Mar 09 '12 at 07:23

1 Answers1

1

If you have a time series of images, I recommend the CONDENSATION algorithm for the interpolation of curves (see also the wikipedia page). There are a number of implementation available, notably in opencv.

meduz
  • 816
  • 6
  • 15