[Insight-users] Curvature Flow - problem

[Luis Ibanez]

Hi Rodrigo,

Thanks for posting the rendering of the image,
it makes things much clearer.

This doesn't look like artifacts due to the
time step. I'm wondering if the problem is
related to the pixel type of your image.

Could you please post some details on the
typedef you are using. Or post a code snippet
from you program, including the image type
and the filter type ?

You could also take the example

   Insight/Examples/Filtering/
       CurvatureFlowImageFilter.cxx

change the Dimension in lines 95-96
and run it on your image. Note that this
example is casting the images to pixel
type float before running the curvature
flow image filter.

Please let us know what you find.

Thanks


    Luis



-------------------------
Rodrigo Trujillo wrote:
> Hi Luis,
> 
> I made what you  wrote, but the results had continued wrong.
> I applied the filter without TimeStep, and later with a lot
> of values. Everytime 3D image were wrong.
> 
> I am sending the 3D image without and with the filter.
> I'm working with ITK 1.4.
> 
> Thanks,
> 
> Rodrigo Trujillo
> 
> 
> 
> ----- Original Message ----- 
> From: "Luis Ibanez" <luis.ibanez at kitware.com>
> To: "Rodrigo Trujillo" <rodrigo.trujillo at cenpra.gov.br>
> Cc: "ITK Users" <insight-users at itk.org>
> Sent: Tuesday, February 03, 2004 4:05 PM
> Subject: Re: [Insight-users] Curvature Flow - problem
> 
> 
> 
>>Hi Rodrigo,
>>
>>This filter has been improved in order
>>to compute an appropriate time-step
>>internally. You no longer need to
>>specify a time step value. (Assuming
>>you are using ITK 1.4 or 1.6).
>>
>>If nevertheless, you decide to play with
>>this parameter, you may want to start
>>using time steps values as low as 0.005
>>for 3D images. Then progressively experiment
>>raising this value as much as you can withouth
>>triggering the numerical instabilities.
>>
>>The tradeoff is between the size of the
>>time step and the number of iterations
>>that you have to apply in order to
>>reach a certain degreee of smoothing.
>>
>>
>>
>>Regards,
>>
>>
>>    Luis
>>
>>
>>
>>------------------------
>>Rodrigo Trujillo wrote:
>>
>>
>>>Hi,
>>>
>>>I´m working with MRI  and testing some filters. When i was testing
>>>Curvature Flow with 2D image
>>>the filter seems to be correct, but when I make the volume, the volume
>>>is wrong.
>>>
>>>The problem is the same when value of TimeStep is set wrong, but I used
>>>the values from "ITK
>>>Software Guide" ( SetTimeStep(0.0625) for 3D images) and others (0.25 ,
>>>0.125 , 0.3125...).
>>>
>>>Which the correct value? What is the problem?
>>>
>>>
>>>Thanks,
>>>
>>>Rodrigo Trujillo
>>
>>
>>
>>
>>
>> ------------------------------------------------------------------------
>>