detecting edges automatically with line selection

[Thomas Theelen]

Dear all,

I want to do the following:
1. Segment OCT images by automatically detecting different gray
value areas (I think this works best with edge detection?)
2. Mark these borders with lines, following the border contour
3. Measure the distance between lines to get tissue thickness
curves
4. Apply the thickness curves to several images in a stack to get
thickness maps or 3D reconstructions of layers
This seems very complex. I would be happy for advice for at least points
1-3.
Linked is a typical image. Analysis should take place in 8-bit.

Link to foto:
http://picasaweb.google.nl/lh/photo/ouTkCBXor7pY-YOGYsDR2Q?authkey=5XvSk
ES-bF4&feat=directlink

Many thanks, Thomas

____________________________________________

Thomas Theelen, M.D.
Radboud University Nijmegen Medical Centre
Department of Ophthalmology (409)
Philips van Leydenlaan 15
6525 EX Nijmegen, The Netherlands
Phone: [phone]
Fax:    [phone]






Het UMC St Radboud staat geregistreerd bij de Kamer van Koophandel in het handelsregister onder nummer 41055629.
The Radboud University Nijmegen Medical Centre is listed in the Commercial Register of the Chamber of Commerce under file number 41055629.

[karo03]

Hi
Some ideas for 1-3:
Segment by threshold the different grey regions (not only the edges),  
calculate from each region the distance transform ("Process -> Binary -
 > Distance Map") the skeleton ("Process -> Binary -> Skeletonize").  
The distance times 2 plus 1 at each location of the skeleton, is a not  
to bad estimate of the thickness in unities of pixels.
Karsten

Am 22.01.2009 um 10:15 schrieb Thomas Theelen:

> Dear all,
>
> I want to do the following:
> 1. Segment OCT images by automatically detecting different gray
> value areas (I think this works best with edge detection?)
> 2. Mark these borders with lines, following the border contour
> 3. Measure the distance between lines to get tissue thickness
> curves
> 4. Apply the thickness curves to several images in a stack to get
> thickness maps or 3D reconstructions of layers
> This seems very complex. I would be happy for advice for at least  
> points
> 1-3.
> Linked is a typical image. Analysis should take place in 8-bit.
>
> Link to foto:
> http://picasaweb.google.nl/lh/photo/ouTkCBXor7pY-YOGYsDR2Q?authkey=5XvSk
> ES-bF4&feat=directlink
>
> Many thanks, Thomas
>
> ____________________________________________
>
> Thomas Theelen, M.D.
> Radboud University Nijmegen Medical Centre
> Department of Ophthalmology (409)
> Philips van Leydenlaan 15
> 6525 EX Nijmegen, The Netherlands
> Phone: [phone]
> Fax:    [phone]
>
>
>
>
>
>
> Het UMC St Radboud staat geregistreerd bij de Kamer van Koophandel  
> in het handelsregister onder nummer 41055629.
> The Radboud University Nijmegen Medical Centre is listed in the  
> Commercial Register of the Chamber of Commerce under file number  
> 41055629.

... [show rest of quote]

Karsten
[hidden email]