Jump to content

Narrowband

Get higher resolution project


Recommended Posts

Hi,
I have struggled with the Drizzle technique many years but had always problem to see any difference when taking images with my teleskop, shorter focal lenses has worked good.

Now with faster computer, Windows10 64-bit and new updated software it works as I expected it from the beginning. The idea come already when I took my first digital astrophoto, the comet Hyakutake 1996. I read about NASA and how they use Drizzling on Hubble images to increase the undersampled images from it.

I have documented my workflow here and a first test with the Globular Cluster M13 as a test object.

http://www.astrofriend.eu/astronomy/projects/project-get-higher-resolution/01-get-higher-resolution-introduction.html

My astrophotos with the telescope are taken from the balcony and never of very high quality, but this time it worked.

/Lars

Link to comment
Share on other sites

I briefly looked at your example and have a few objections, if you don't mind.

Putting aside if new workflow works better for you and you like images made with such workflow better - that is just fine but it's not relevant to drizzle or topic if drizzle works.

My position is that in amateur setups - drizzle does not work, and is in fact waste of SNR.

In order to test this, I advocate following workflow - use same software for alignment and stacking, instead inspecting images after processing - do analysis on what is actually relevant - compare star FWHM in drizzled and non drizzled image. In fact - throw in control as well. Here is complete workflow:

Calibrate your images in AstroimageJ as you usually do and then do split demosaic (extract two green fields out of the image first field will have odd rows/columns, while other even- let's use monochromatic data for the test - it is easier to do). Both green fields will have half the original resolution - but don't resample them or interpolate them - that is actual sampling rate.

Stack original data without modification to one stack - measure FWHM on couple of stars that don't saturate but have good enough SNR

Do drizzle stack x3 to another stack - measure FWHM on same stars but divide result with x3 (because of drizzle factor).

For control - use advanced resampling method like lanczos3 resampling to enlarge each sub prior to stacking - scale them x3 larger, stack normally and again measure FWHM - divide result with x3.

While you are at it - measure signal of one star (AstroimageJ photometry tool) and measure standard deviation on patch of background without stars - divide the two to get SNR for comparison.

Link to comment
Share on other sites

Hi,

All feedbacks are wellcome !

I need a new workflow because the AstroImageJ doesn't have any Drizzle function, maybe I later develope one, but not now. AIJ is more aimed for microscopy. But I still want to use AIJ for the demosaicing and calibration process with my own routines and ideas. And what I do is exactly what you write, the two green channels must always be handled separately, even when you not do Drizzling.

Drizzle always need more exposure because you let the photons that's normally detected by one pixel be divided on 9 new pixels (in this example), and thus you need 9 times more exposure. If I can get my heavily undersmapled images to be of more high resolution by just increasing the number of images it's a low price to pay. Of course I can't compete with professionals, but if I think like that I don't need any telescope either, I just have to look on internet and Hubble images. But no fun for me.

 

I noted from earlier post from you that you have very high knowledge of physics/math, do you work with some astrophysics or nearby ?

 

ps.

I will do measurments on the images, but not now, now I'm very curious to see what I can do with my old photos.

 

/Lars

 

Link to comment
Share on other sites

Hi,

No, my profession is programming / system architect - I do physics and astronomy as a hobby.

I can do above example and post here results. In fact, I think that I already did something similar once - to show to people that drizzle is not doing what they believe it is doing. I just downloaded latest version of DSS and I have some old data that I can use. It is not under sampled but can be easily made so by software binning.

In your example with globular cluster - you enlarged regular version by x3. What interpolation did you use for that? Much depends on interpolation algorithm.

Also - looking at the stretched images can be misleading. We tend to stretch as far as image looks pleasing. If two images have different SNR - they will have different level of stretch as a result.

These are just some reasons why you should not visually judge actual resolution of the image. I'll quickly demonstrate both points above.

image.png.c2cfecc940d72330776edec28e6e6a5f.png

This image shows under sampled star being enlarged by two different algorithms. Same base under sampled image was used - left image is bilinear interpolation, right image is Cubic O-Moms. Star in the right image looks "tighter" and better resolved than left

Here is example for other thing that I mentioned above:

image.png.7369f9d4bf9378ebbb4be8e0461a1012.png

Which star do you feel is "tighter" and which image has "better resolution" - left one, right? star is smaller. In fact - it is the same image, left one polluted with x3 higher noise than right - making SNR x3 larger in the right image (drizzle x3 will produce x3 less SNR) and I applied linear stretch to both until background looked the same.

Since we have point sources in the image - stars, we do have PSF for our linear images - and that is what we should measure because that is perfect measure of resolution of the image. Visual clues can be misleading.

Link to comment
Share on other sites

Hi,

I know there are a lot of problems when comparing two images, but I have tried to do the same process on both. I used IrfanView to interpolate the two images, only to have them at the same size. IrfanView is configured to use the Lancoz algorithm.

 

Drizzling images taken with my TS130 APO telescope has earlier always been hard and no obvious increase of the resultion. Maybe because I have terribly conditions on my balcony. Photos taken with my mobile equipment which I can take out to dark places works much better. I'm working on a Sigma APO 150 mm f/2.8 image now, still the M13 object.

http://www.astrofriend.eu/astronomy/projects/project-get-higher-resolution/03-get-higher-resolution-second-test.html

I'm not finished with it yet, coming soon.

The goal is to both do it mathematically correct and have a pleasing look of the image, as far I can come with my light pollution. Most of the noise comes from the background.

It will come more data later from the 32-bit files, but first I test this on my old images, and try to optimize the Drizzling process and the preprocess. I'm very happy already now how it works, earlier it almost chrashes everytime I did it on a big image.

/Lars

Edited by Astrofriend
Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue. By using this site, you agree to our Terms of Use.