Prev: Problem with adding legend?
Next: fft and ifft : and fftshift
From: Louizi on 21 Feb 2010 13:19
Hi again,
I'm trying to compare between two shapes (proper bone of the nose) : one is either healthy either fractured, the other is healthy in order to categorize the nose I'm treating.
The problem is that the healthy nose cannot be considered as an atlas image (to be compared to) because of the large amount of differences between human noses. So I'm searching for some hints to know which way I will take (decomposing the shapes in different points and comparing based on a certain threshold, zooming or reducing to compare the same sizes of the images..)
Thanks in advance.
Two examples of images : http://www.louizi.com/thesis/1/nose.html
From: michael scheinfeild on 21 Feb 2010 13:56
"Louizi " <mehdi.louizi(a)gmail.com> wrote in message <hlrtem$n26$1(a)fred.mathworks.com>...
> Hi again,
> I'm trying to compare between two shapes (proper bone of the nose) : one is either healthy either fractured, the other is healthy in order to categorize the nose I'm treating.
> The problem is that the healthy nose cannot be considered as an atlas image (to be compared to) because of the large amount of differences between human noses. So I'm searching for some hints to know which way I will take (decomposing the shapes in different points and comparing based on a certain threshold, zooming or reducing to compare the same sizes of the images..)
> Thanks in advance.
> Two examples of images : http://www.louizi.com/thesis/1/nose.html

you need to look material about shape descriptor.
you have function in matlab region props.
this can describe shape.
try area , major and minor axis , euler number , .
you can use fourier descriptor of shape [ take edges make them f=x+j*y and make fft]
also find centroid and make shape number , you can get diffrential chain code of the shape.

after you comupte feuers of the shapes you need to classify the shapes
use linear classifier,optimal classifier , neares neighboor or svm[support vector machines]
have good luck !!
From: Louizi on 21 Feb 2010 14:28
Thansk a lot for your quick reply and for your help. I have enough material to dig now!
 | 
Pages: 1
Prev: Problem with adding legend?
Next: fft and ifft : and fftshift