Back to forums

Image that is undersampled already too large to drizzle

ZWO ASI071MC COOL Acquisition Drizzle Deep Sky Help Needed
6 replies•238 views
Frank Szabo avatar
Frank Szabo avatar
I got an issue I can't fix, don't know how. 

I recently took a 2 panel mosaic of the veil nebula. Very short integration- but that's not the point. 



The problem is, that this camera 071mc pro has a very large micron size if you ask me.  I mean sure it's fine with a high power refractor or SCT like 1200mm, but like  less than 600mm focal length  it's just not a favorable pixel size.   (zwo071mc pro = 4.78micron pixel pitch) 
I end up drizzling most of the time, because otherwise I end up with square stars. 
But just a few days ago I did this mosaic, and the picture size is already over 70megabytes  (png format) .   Drizzling this would increase the size to something huge, I can only save it as an 8bit .jpeg, which I really really don't like lately.  Sure, fine for social media, but to save it for myself or other sites, it's not. 

So I even tried to upload it here, and I don't really want to upload a 70 + megabyte image either. 

Does anyone have any idea how to deal with undersampled  images? 

 I know the drizzle, but as I said, it will be humongous image and then I have to resample it anyway.  

My other option is buy a different camera?   Use it on a higher focal length telescope only? 

image 1: original size (.png)
imae 2 : resized down 50% - note the stars are blocks. 



See, some screenshot here if I would just resample it 50% not how the stars are now jus blocks,
Ashraf AbuSara avatar
Ashraf AbuSara avatar
Sorry might be a silly question, but why do you care how large the image is? Storage is pretty cheap at about $10-$15/terabyte these days for external storage. It might take a while to process though. Drizzle 2x. process it, and resample it, and you can free up the data after you are done. I upload images here to Astrobin in lossy Jpeg format, but it is still pretty good. 

Some of my fits files working with a full frame sensor at 3.76um can go upward of a 1000mb.
Supportive
dkamen avatar
dkamen avatar
If you increase resolution by a factor, image size increases by the square of that factor. Doesn't matter if you drizzle, use a larger scope or a camera with smaller pixels.  Of course each solution has different pros and cons and the physical ones are better than drizzle. But size-wize they are the same. Fact of life. 


FWIW in the screenshot you have resampled 500%, not 50%.
Helpful Concise
Frank Szabo avatar
Frank Szabo avatar
If you increase resolution by a factor, image size increases by the square of that factor. Doesn't matter if you drizzle, use a larger scope or a camera with smaller pixels.  Of course each solution has different pros and cons and the physical ones are better than drizzle. But size-wize they are the same. Fact of life. 


FWIW in the screenshot you have resampled 500%, not 50%.

I have zoomed in on the image to see the star shapes in Photoshop, it's not resampled at 500X.  What you see 500x is the zoom level.  The image is sampled to 50% original size.
Kyle Goodwin avatar
Kyle Goodwin avatar
If you downsample it the stars are going to look blocky, you're essentially undersampling again an already undersampled image.  I'm not understanding the comment about PNG vs. JPG, do you just mean for posting here?  I post ~100MP images here without any issue.  A JPG that size is usually about 75MB.  It's not a problem.  For my own archival I'm not concerned about the JPG, though, I'll keep my final .xisf or .fit and I also keep the raw subs and calibrations frames in case I want to reprocess it or combine it with new data in the future.  I've got 12 years of data.  Storage is cheap.
Helpful Insightful
Quinn Groessl avatar
Quinn Groessl avatar
It's probably not best practice, and probably hurts the final quality a bit, but I take mine in to mspaint sometimes and resize them there.
Wei-Hao Wang avatar
Wei-Hao Wang avatar
Several points.
  1. What's the evidence that a 4.8 micron pixel would under-sample the stars from your 600mm scope?  Did you measure a FWHM that's smaller than 2 pixels in your stacked images or see square stars?  Or it is just some random smart website that tells you your stars would be under-sampled?  If it's the latter, forget about that website and never visit it again.

    2. A 4.8 micron pixel under a 600mm focal length means a pixel size of about 1.6". Under most seeing conditions for amateur (2" to 3"), this can mean under sampling. However, at focal lengths under 1000mm, seeing is not the only thing that determines the star size on the focal plane. The size is strongly affected by aberration, and tracking error can play a role as well. You should only consider your image under-sampled if your measurement tells you it is under-sampled. 

  2. Even if your 4.8 micron pixel does not under-sample the stars, the star FWHM in your stacked image can be larger than what it should be after stacking. This is because the 9.6 micron Bayer pattern can easily under-sample the stars (seeing plus aberration) especially if your optics is excellent. In such a case, what you really need is not the regular drizzle (which increase the image dimensions by 2x or more). You need Bayer drizzle. It will not increase the image size, and can still improve sharpness.

  3. We need properly sampled stars if we want to apply filtering effect that works at pixel scales, such as deconvolution and sharpening, and perhaps even noise reduction. Such algorithms do not work well if your stars are under-sampled or are even squire-shaped. Unless you heavily rely on such, slightly under-sampled stars are fine. (Severe under-sampling that leads to square stars is not a good thing of course.)

  4. Down-sampling the image before uploading to internet for people's viewing is a good thing to do. No one needs to see your full-resolution images or to see stars that are 3-pixel across. Such level of details is good for your own processing and printing. On the other hand, it is a waste of network resource and it does not provide useful additional information to viewers. As long as the stars remain round (or even a tiny bit of squareness) at 100% viewing, it is good enough for viewers on the internet. This is particularly true if your image is a wide-field mosaic. Using your veil nebula image as an example, I clicked into it and checked the full resolution image. I would say the viewer experience would not be very different if you down-sample it by 1.5x or even 2x.The last screenshot you attached does not look good, of course. But who will look at your image at 561% enlargement?  No one.


My 2 cents.
Helpful Insightful