[Insight-users] Re: reg 3D registration

[Luis Ibanez]

Hi Prasad,

I assume that you are doing this with the 64x64 images
you recently constructed.

If this is the case, a step lenght of 4 is extremly
large. This is the step size that we will normaly use
for a 512x512 pixels images.  Note that actually the
step lenght is given in millimeters, so, it is *very*
important to be aware of the pixels spacing in your
image.

What is the pixel spacing in millimeters in your images ?


I would suggest you to start with a step size of 0.2 or so,
and still set the minimum to 0.01.

A simple exercise that will give you an indication of
how apropriate is your step size, is to plot the metric
values along the iterations. This is done in all the
registration examples of the SoftwareGuide.pdf.

If the plot is noisy and has multiple sudden jumps, that
usually indicates a too large step size.

Please take a look at the plots in the "Registration" chapter
of the software guide, so you can have an idea of how a metric
plot should look like.



Regards,


   Luis



-------------------------

Prasad Sarma wrote:
> hi Luis,
> 
>    Thanks alot for all your prior replies.I have
> attempted to do registration and I have confronted the
> problem with the command/observers too.
> I changed the parameters in registraion3.cxx to
> include dimension 3 and  pixel type to be unsigned
> char and also changed the parameters for traslation to
> include 3rd dimension.
> I changed the step length and the no of iterations but
> my efforts were in vain.
> 
> for
> the max step length is 4
> min step length is .01
> max iterations: 200
> for RegularStepGradientDescentOptimizer,liner
> interpolation, translation transform
> (same as registraion3.cxx)
> 
> here's the o/p
> 
> ---------------------------------------
> C:\Documents and Settings\Prasad\My
> Documents\Registerbin\Debug>Register.exe 
> 
> My3Dimage_1.mha My3Dimage_2.mha
> 
> 0 = 1441.89 : [2.2464, 1.88965, 2.71715]
> 
> 1 = 1703.4 : [3.64359, 3.27332, 6.20044]
> 
> 2 = 2347.09 : [2.36144, 2.56046, 4.84106]
> 
> 3 = 1955.97 : [0.7978, 3.77632, 4.56411]
> 
> 4 = 1830.46 : [1.66003, 3.79176, 5.07039]
> 
> 5 = 1982.66 : [2.64556, 3.85917, 5.22589]
> 
> 6 = 2115.72 : [3.43758, 3.57994, 4.68298]
> 
> 7 = 2179.97 : [3.03342, 3.73681, 4.93208]
> 
> 8 = 2283.28 : [2.60642, 3.65857, 4.68398]
> 
> 9 = 2032.24 : [2.77633, 3.7366, 4.84994]
> 
> 10 = 2135.4 : [3.00294, 3.76909, 4.95041]
> 
> 11 = 2302.2 : [2.90712, 3.73601, 4.87727]
> 
> 12 = 2187.98 : [2.79079, 3.72264, 4.83351]
> 
> 13 = 2130.96 : [2.83514, 3.74182, 4.87315]
> 
> 14 = 2163.11 : [2.82722, 3.72772, 4.84642]
> 
> 15 = 2147.48 : [2.83501, 3.73382, 4.8585]
> 
> 16 = 2155.71 : [2.84145, 3.7403, 4.87119]
> 
> 
> abnormal program termination
> 
> C:\Documents and Settings\Prasad\My
> Documents\Registerbin\Debug>
> 
> ----------------------------------------------
> 
> will this error arise due to the improper setting of
> the parameters while  creating the raw files?
> could you pl give some suggestion?
> 
> Thanks 
> 
> Prasad.
> 
> 
>   
> 
> 
> 
> __________________________________________________
> Do you Yahoo!?
> The New Yahoo! Search - Faster. Easier. Bingo
> http://search.yahoo.com
>