[Insight-users] Re: Better separation of peaks

[Luis Ibanez]

Hi Radhika,

You may want to try running any of the edge-preserving
smoothing filters on your image before extracting the
histogram.

These filters increase the homogeneity of regions and
may help to reduce the variance of the classes associated
with particular tissues in your image.  

You may want to try filters like:

   itkMedianImageFilter
  itkCurvatureFlowImageFilter
  itkGradientAnisotropicDiffusionImageFilter
  itkCurvatureAnisotropicDiffusionImageFilter


Markov random fields could also help, but they may be
an overkill if the purpose is just to enhance the peaks in
the histogram.

Since these smoothing filters have parameters, you could
make things more interesting by actually hardcoding
a loop in which the separation criteria of the peaks in the
histogram could drive the parameters of the filters.
In other words, smooth the image until its histogram
satisfy a certain degree of peak separation.

You could use the OtsuCalculator itself as a criterion,
(in that case we will have to modify it a bit, in order
to return the value of the inter-class variance that is
already being computed internally....


Regards,


     Luis


----------------------------------
Radhika Sivaramakrishna wrote:

> Hi Luis,
> I have a set of T2 weighted brain MRI images. I would like to separate 
> background from tissue. In some cases, there are clearly two peaks in 
> the histogram whereas in other cases, I cannot clearly distinguish two 
> peaks [the histogram looks quite uniform], however, when I examine the 
> image, the tissue is quite well separated from the background. I was 
> wondering if you could recommend any steps to actually make the 
> histogram more "peaky" so that I could use some histogram parsing 
> techniques to separate background from tissue.
>
> Thanks
> Radhika
>
>                                                     
> ----------------------------------------------------- 
> Confidentiality Notice.
> This email message is for the sole use of the intended recipient(s) 
> and may contain confidential and privileged information. Any 
> unauthorized review, use, disclosure or distribution is prohibited. If 
> you are not the intended recipient, please contact the sender by reply 
> email and destroy all copies of the original message. If you are the 
> intended recipient, please be advised that the content of this message 
> is subject to access, review and disclosure by the sender's Email 
> System Administrator.
>