Welcome to iraf.net Friday, March 29 2024 @ 01:37 PM GMT


 Forum Index > Help Desk > Applications New Topic Post Reply
 Inaccurate FWHM of output PSF from PSFMATCH
   
rdcschan
 08/29/2013 01:29PM (Read 2783 times)  
+----
Newbie

Status: offline


Registered: 01/21/2011
Posts: 5
Hi everyone,

I am trying to match the PSF in different HST/WFC3 bands, but no matter which kernel size, grid size, and filter options (including frequencies) I used, I cannot get the FWHM of the output PSF in SOME bands correctly.
The output PSF are somewhat wider (>5%) compared to the reference PSF, and occasionally show negative spikes at regions offset to the stars. I did not use Daophot to select stars, but used as stacked PSF images in individual bands as input, and the widest stacked PSF as reference image, although I have tested that using DaoPhot will give me the same result.

Thank you very much.JeffreyP.S. A quick search in this forum showed a similar post before, but unfortunately no answer is given.

 
Profile Email
 Quote
valdes
 08/29/2013 01:29PM  
+++++
Active Member

Status: offline


Registered: 11/11/2005
Posts: 728
Hi,PSF matching is a tricky process. I have not used the PSFMATCH task but I could try and see if there is anything you are missing. Please send the parameter settings, at least for what you think should be the right settings. However, it is most likely this is a data dependent problem which I would not be able to judge.My off-hand guess is that there is a registration issue.Good luck,
Frank Valdes

 
Profile Email
 Quote
rdcschan
 08/29/2013 01:29PM  
+----
Newbie

Status: offline


Registered: 01/21/2011
Posts: 5
[quote:a598b51796="valdes"]Hi,Please send the parameter settings, at least for what you think should be the right settings.
My off-hand guess is that there is a registration issue.Good luck,
Frank Valdes[/quote:a598b51796]Hi Frank,
Thank you for your reply. Sorry for my late response, as I made some quick images to simplify the situation so everyone can understand the situation more clearly:[img:a598b51796]http://i.imgur.com/xRQImwi.png[/img:a598b51796]
Suppose I am trying to match PSF A & B to this reference PSF, which has the broadest FWHM... I have to do more bands but this simplify the situation. There are stacked PSF from the images so the registration should be good enough, however the S/N is not very high since the field I am dealing with is not very large.Here are the general parameters I used, actually I did a grid search of the parameters dnx/dny , pnx/pny (from 3 to 49) and sx2/sy2 (from 50 to 6000):# TASKMETA: task=psfmatch package=immatch
input: '@psf_rmstackin4.list'.....'Input images'
reference: '@psf_rmstackref4.list'.....'Reference images or reference psfs'
psfdata: 'psf_rmstackstar.list'.....'Objects lists or input psfs'
output: '@psf_rmstackout.list'.....'Output convolved images'
convolution: 'image','|image|psf|kernel|',,'Kernel computation method'
dnx: 49.....'X width of data region to extract'
dny: 49.....'Y width of data region to extract'
pnx: 5.....'X width of convolution kernel'
pny: 5.....'Y width of convolution kernel'
center: yes.....'Center the psf objects ?'
background: 'median'.....'Background fitting function'
loreject: INDEF.....'Low sigma rejection threshold'
hireject: INDEF.....'High sigma rejection threshold'
apodize: 0.0.....'Fraction of endpoints to apodize'
fluxratio: '1'.....'The reference to input integrated flux ratio'
filter: 'cosbell'.....'Filter/replace option'
sx1: INDEF.....'Inner x spectral frequency for cosine bell filter'
sx2: 750.0.....'Outer x spectral frequency for cosine bell filter'
sy1: INDEF.....'Inner y spectral frequency for cosine bell filter'
sy2: 750.0.....'Outer y spectral frequency for cosine bell filter'
radsym: no.....'Radial symmetry for cosine bell filter ?'
threshold: 0.2.....'Threshold in fourier spectrum for modeling/replacing'
normfactor: 1.0.....'The kernel normalization factor'
boundary: 'nearest','|constant|nearest|reflect|wrap|',,'Boundary extension'
constant: 0.0.....'Constant for constant boundary extension'The result below use the parameter listed here, except varying the kernel size (pnx/pny)Let's first see the acceptable one - PSF B,
The image below shows the matched PSF B, with different panel showing the result of using various kernel size (pnx/pny):
[img:a598b51796]http://i.imgur.com/tm8BGN8.png[/img:a598b51796]
As you can see, the FWHM matches exactly to the reference PSF (2.41). For small kernel sizes, there appears some negative spikes around the star, but you can see the PSF quality improves as a larger kernel is used. For PSF A, the situation is much worse.
The image below shows the matched PSF A, with different panel showing the result of using various kernel size (pnx/pny):
[img:a598b51796]http://i.imgur.com/shuaTHN.png[/img:a598b51796]
No matter what kernel size I used, the FWHM are much larger (~2.55) compared to the reference. Moreover the quality of the PSF are terrible, with numerous spikes at the boundaries. Even for very big kernel sizes (up to 37 x 37 pix), you can still see positive and negative residuals.
I wonder if there is a parameter I overlook or something I should do to prevent this. Thanks again!

 
Profile Email
 Quote
valdes
 08/29/2013 01:29PM  
+++++
Active Member

Status: offline


Registered: 11/11/2005
Posts: 728
Hi Jeffery,Thanks for the pictures, that helps understand things better. It looks like there are edge effects. The only thing I see to suggest, if you haven't played with it already, is the apodize parameter. Other than it is a fraction and so should be less than 1, I don't know what values to suggest. As part of this edge concern, it's noted that background has to be fairly accurately subtracted to work well. So making the background far away (to avoid the diffraction spikes) and working with the background parameters (or externally doing background and using "none") may be important.Just some ideas to consider. None of this directly touches on the final FWHM that you are concerned about.Frank

 
Profile Email
 Quote
rdcschan
 08/29/2013 01:29PM  
+----
Newbie

Status: offline


Registered: 01/21/2011
Posts: 5
[quote:22649eebc5="valdes"]The only thing I see to suggest, if you haven't played with it already, is the apodize parameter.
So making the background far away (to avoid the diffraction spikes) and working with the background parameters (or externally doing background and using "none") may be important.[/quote:22649eebc5]I see, I haven't played around with the apodize parameter yet, let me include it in the grid search to see if I can get out these edge thing.... For the background as the background is already removed during the stacking process, but it is worth to try with none too. I will let u know how it goes.Thanks again

 
Profile Email
 Quote
rdcschan
 08/29/2013 01:29PM  
+----
Newbie

Status: offline


Registered: 01/21/2011
Posts: 5
Hi, I just played around with apodize with the other parameter I used above, but with background = 'none';It turns out that the background parameter have no observable difference at all, probably due to the fact that the background is already subtracted. On the other hand apodize parameter gives me these:[img:fa2faeef91]http://i.imgur.com/TZLz3cd.png[/img:fa2faeef91]The pattern of the boundary splikes changes as I used different apodize parameter, but it seems they still refuse to go away Cry
I m officially run out of ideas at this point, I would appreciate if someone can share his experience on such kind of work.

 
Profile Email
 Quote
   
Content generated in: 0.14 seconds
New Topic Post Reply

Normal Topic Normal Topic
Sticky Topic Sticky Topic
Locked Topic Locked Topic
New Post New Post
Sticky Topic W/ New Post Sticky Topic W/ New Post
Locked Topic W/ New Post Locked Topic W/ New Post
View Anonymous Posts 
Anonymous users can post 
Filtered HTML Allowed 
Censored Content 
dog allergies remedies cialis 20 mg chilblain remedies


Privacy Policy
Terms of Use

User Functions

Login