| Completing a Gaussian Fit [message #57483] |
Wed, 12 December 2007 14:02  |
rpertaub@gmail.com
Messages: 43
Registered: January 2007
|
Member |
|
|
Hello,
I have a problem I am not sure how to go about...more of a physics
question maybe than IDL, but does not hurt to ask...
I have an image of a spot. The spot can be assumed to be fairly
Gaussian, i.e bright in the middle and dissipating as radius
increases. All is good. Except say now I have only part of the spot.
(Say if radius is 30, I have only a spot of radius 5). How do I model
a Gaussian based on only this information and nothing else? I have no
idea the intensity at the std deviation point, or radius 15? Is it
possible? How does IDL do Gaussian fit?
Thanks,RP
|
|
|
|
| Re: Completing a Gaussian Fit [message #57574 is a reply to message #57483] |
Thu, 13 December 2007 01:21  |
Bringfried Stecklum
Messages: 75
Registered: January 1996
|
Member |
|
|
rpertaub@gmail.com wrote:
> Hello,
> I have a problem I am not sure how to go about...more of a physics
> question maybe than IDL, but does not hurt to ask...
>
> I have an image of a spot. The spot can be assumed to be fairly
> Gaussian, i.e bright in the middle and dissipating as radius
> increases. All is good. Except say now I have only part of the spot.
> (Say if radius is 30, I have only a spot of radius 5). How do I model
> a Gaussian based on only this information and nothing else? I have no
> idea the intensity at the std deviation point, or radius 15? Is it
> possible? How does IDL do Gaussian fit?
> Thanks,RP
This is fairly easy. You just need to apply a mask on your model image
when computing the chisquare during the minimization, i.e. something like
mask=(object gt sigma*noise)
chisq=mask*(object-model)^2
Of course the model parameters will be more uncertain if the observed
fraction of the Gaussian image becomes smaller. In other words, you need
fairly high signal-to-noise to get meaningful results.
regards,
B. Stecklum
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
