Super resolution from stacked images?

[CrimsonSkies]

I have seen high res astro images made from multiple images (even video) and am trying to do something similar.

Shooting in RAW I can make 10+ images of a scene really easily, but I cant figure the stacking workflow to improve resolution of the final image

Using Astro stacking doesnt seem to create an image better than a single shot, though I could be doing it wrong

[NotMyFault]

Hi,

stacking in Affinity can be used for different purposes. Super Resolution unfortunately cannot be achieved.

You need a Camera capable of sub-pixel shift between exposures, and a Camera specific Software to merge these images  

With Affinity Photo, you can stack for:

• Focus Merge
• HDR Merge
• Astrophotography Merge
• Stacking with automatic image alignment 
• Stacking to reduce noise

[CrimsonSkies]

Awesome, thanks for the reply - now I understand a bit more of what's going on, dont have a camera that clever  

[Suburbian]

Super Resolution "can" be achieved ....and with any camera, not just on Photoshop, who if you have a system with the resources or an Apple M1 chip is "awesome"  but also with Stacking On Affinity as well, it's been done and it's been posted on this forum starting back in 2017

 

Don't take my word for it I'm just a Troll? 

 

The author of that post proved it and several members verified it as working following the steps outlined

 

I used the Photoshop system for taking older 10 meg D200 images and increasing the resolution, with just built in graphics and 8 gig of RAM 

on a Dell, I could process one shot and have to reboot the system again before I could try another or it would lock up 

 

You're going to be maxing out resources quickly, no matter what you try 

 

But don't let ANYONE ever tell you "NO" and just accept it is as Final 

 

That's what caught my attention the most, how someone (anyone) could come along and say, "unfortunately no" without a single thing to support it and the person who asked just accepted it as Final? 

 

\

 

A troll who says, Search "Super Resolution Using Affinity" on Google and see what pulls up? More than just my word for it 

 

If you want to use Adobe's Superresolution, this Troll highly suggest you have a one good system, Apple's M1 will allow you incredible resolutions and everything else will struggle, maybe not right away but they will fade and crash, you'll be amazed and then po'd you don't have more resources... 

[John Rostron]

You can approximate the sub-pixel shift by taking a series òf images hand-held. I used my camera 's equivalent of motor drive. I then enlarged each image twofold, stacked the resultant images with align ticked and took the mean or median.

I have to say that the results were not noticeably better than the twofold enlargement of any one single image.

John

 

[Suburbian]

Approximate? 

Close to the actual, but not completely accurate or exact ...the definition of "Approximate" aka; a Complete Stab in the Dark at Something ? 

Here's a fact about pixel shift (researchable as well on this very forum?) it can be "accurately" produced by the human hand. But it's not my word, it happens to be the word of a person here on the forums, it's "published" and other members reported duplicating the results themselves, publishing their opinions. Not one or two people, several. 

it's cultural thing I suppose, in some societies free speech means you're free to write whatever the heck you want, facts are meaningless only your right to say anything is, right or wrong doesn't matter.

 

imaginary tales are the ones people tell with no steps attached to them for us to follow on our own, because they never existed in the first place 

 

the secret kingdom of "I" 

 

Superresolution means Super Flaw Reveal, count on it

 

It really can boil down to how much time and patience you have or how consistently stable your hands are...if you take 20 images at one time handheld, some are going to be trash,   don't use auto WB , aperture or shutter priority, use all manual controls all the way ... if you don't nail your settings and sort the 20 images? What you get is 20 images of varying flaws, all multiplied when stacked and one bad apple will spoil the whole bunch ...but not always? 

[vickosis]

Indeed it's possible to take a set of similar images e.g. burst of handheld images taken in quick succession (mimicking pixel shift as @John Roston rightly points out). I've done this in the past with low night time photography using entry level APS-C camera with high ISO settings and the result was quite impressive in terms of resolution and noise reduction, although I have to confess I did use my previous Photoshop subscription to perform the stack but it can also be done with Affinity Photo as I just found out.

The thing I struggle with tho is the difficult to resize images using % percentage, which isn't one of the options for some reason and to allign images you need to use a new stack instead of just being able to do this with existing layers which is a pain IMO as it's not an great workflow. To be fair I've only tried this just now for the first time and what lead me to this post was a Google search of "super resolution stack  +affinity +photo" similarly to what @Suburbian suggested. Was quite surprise to come across this thread and how quick some peeps are to dismiss this as not possible, but perhaps what they mean is there is no built in Affinity method for this specifically.

As this is a fairly common task for peeps like me it would be great if Affinity could perhaps look at simplifying this process and adding these in a future release (pretty please)
 

[NotMyFault]

There is probably some misunderstanding or cross-talking.

Of course you can improve image quality a bit by stacking with Affinity. 

• you can reduce noise
• you can increase color bit depth
• You can partially increase resolution up to the actual image resolution in case single images are blurred and have effective resolution lower than document size. This is important as Bayer Filter always gives lower effective resolution for color images. You can boost up to actual document resolution. 

what does not work is up-sampling images and getting e.g. 2x resolution from multiple time mages of 1x resolution in Affinity.  This process / algorithm is not implemented in Affinity, but some Camera vendors have implemented it in Camera (Pentax) or PC software (Sony)

Feel free to prove me wrong. For a prove we need:

• a set of input images with defined size / resolution, eg 24 Mpx, e.g by limiting RAW resolution (Canon can do this)
• a workflow description (list of steps, video tutorial, macro) 
• export image from workflow with 2x resolution, eg 48 Mpx
• a reference image shot with a Camera / lens capable to provide 48 Mpx (raw file) to compare

 

 

[vickosis]

@NotMyFault interesting, but why do I see so many articles claiming that it's possible to mimick pixel shift of static subject by using a burst of images handheld and stacked in post?

In example ...

https://petapixel.com/2015/02/21/a-practical-guide-to-creating-superresolution-photos-with-photoshop/  or
https://www.dpreview.com/articles/0727694641/here-s-how-to-pixel-shift-with-any-camera

PS IMO The benefits you listed alone that improve image quality are good enough resons to use stacking at times 😊

 

[NotMyFault]

Because I tested those recipes on my own and found that while these methods improve sharpness a bit it will never effectively give 2x resolution. Calling this methods "super resolution" is misleading in the same way as Tesla calls its cruise control "autopilot". It is marketing bullshit.

For a hardcore test you would need to take images of high frequency test patterns, and prove that this stacking will actually deliver higher frequency results.

I have never seen such factual proofs, only casual test images which look a bit better.

 

Once again for clarity:

• I fully agree that stacking can improve visual quality and perceived sharpness. I have used it for that purpose.
• I'm just not convinced that you can gain 2x upscaling in the physical sense of getting small details like a strict black / white checkered board of 1px size. For a test, display a checkered bard of e.g. 2000x2000px, and take 4 images with a camera where the image covers only 1000x1000px of the sensor (feel free to scale to other sizes as long as every camera px covers 4 source image px). Then try to super stack these image and compare to the original image. Looking forward to your images. Please upload the 4 images taken, and the AFPHOTO file containing the stack.
• As counter-test: take image of test pattern where every source px is covered by one camera px.
• Note that this test makes most sense when done with a Monochrome sensor camera, to avoid impact of bayer filter. Color cameras actually device only about 50% of advertised resolution.

checkered 256x256 misaligned.afphoto

 

[vickosis]

@NotMyFault I am not exactly disagreeing with you and tbh I have way too much on my plate these days to conduct such extensive tests so I will take your word for it, but at least we do seem to agree that it does improve image quality somewhat.

A while back I did use this "superresolution" method which I linked in my previous comment for some handheld shots of an iconic landmark on my way home around midnight with high ISO and the stack did indeed improve the noise and image quality considerably. Haven't got the image to hand but here it is posted on instagram: instagram.com/p/CZfOB44K3hf