Affiliation
American Association of Variable Star Observers (AAVSO)
Wed, 09/25/2013 - 13:48

Hello

I got this answer from a support ticket from Maxim.  This is in answer to the question of whether Maxim resamples the data when using the "1 star method" with or without "Use Centroid".

 

The "Use Centroid" option for Manual 1-star alignment simply adjusts your clicked point of reference in the image to lock on to the nearest (brightest) centroid. This is done to ensure maximum accuracy in the alignment process. If you leave "Use Centroid" unchecked, it will simply take the points you provided in each image without adjustment for objects in the image. Regardless of this option, the alignment is calculated at the sub-pixel level, and the image is resampled using either bilinear, or bicubic interpolation (depending on which method you specified in the options menu).

 

I would advise not aligning images prior to photometry if you can avoid it, since the auto-matching tool in the Photometry suite has some object detection built in — but if you must align the images, using bilinear interpolation will help minimize the smearing from the resampling process.

 

Sorry for the bad news.  This just goes to show you that there are no stupic questions.

 

 

Gary

Affiliation
American Association of Variable Star Observers (AAVSO)
Alignment methods

Thanks for the detective work, Gary.

FYI - here is a section of MiraPro's help manual:

Resampling method

Specifies the method used to compute image intensity at fractional pixel coordinates.

Bi-linear: Computes a linear interpolation between pixels in the x and y directions.

Bi-cubic: Computes a 4 term cubic polynomial interpolation between pixels in the x and y directions.

Nearest Neighbor: Uses the pixel nearest to the transformed coordinate instead of re-sampling the pixel values. This inflates the FWHM by about 0.3 pixels but preserves the noise structure of the image, which is usually important when Aperture Photometry will be done on the registered images.

The Nearest Neighbor method does preserve the original ADU count in each pixel.  

Richard Sabo
SRIC

Affiliation
American Association of Variable Star Observers (AAVSO)
Does it matter?

The question is "does it matter?"

In an effort to shed some light on that I question I did the following

I took approximately 100 images from a time series I took earlier this year (PNV J19150199+0719471).  The series was almost 500 images and I grabbed a block of about a 100 images that were all on the east side of the meridian. But as it turns out, the airmass was higher than I would have liked but still less than 2.0  The images were taken 1x1 with a V filter.

I then stacked these images 5 images at a time, giving MaxIm plenty of opportunity to slice, dice and mash the photons.

I then measured the 126 star (v=12.625, B-V = 0.961) using the 136 star as the comp star and the 141 star as the check star in both the 100 single images and the 20 five star stacks.  The chart # is 12686CQH.

I also averaged the results 5 stars at a time (running average) to approximate the 5 star stack.

Results:

100 single images                 v=12.6678    sd=0.019

Running ave of 5 images       v= 12.6681   sd=0.011

stacks of 5 images                v=12.6671     sd=0.011

So in this case, out to 3 decimal places there is no difference between analysing the images unstacked and stacked 5 at a time.  In addition a decrease in sd is about what I woud expect from experience...something less than the theoritical 1/sqrt(N).

There is a difference of 0.001 beteen the stacked images and the 5 star running average.  The sd's are identical to 3 deecimals.  If you are uneasy about stacking just average the magnitudes to reduce the error.

This test obviously isn't despositive but is bette than no analysis at all.

And yes I know that my measurement was approximately 0.04 mag dimmer than the comp star mag.

Jim jones