Jump to content
You must now use your email address to sign in [click for more info] ×

Recommended Posts

I have spent much of the morning with a batch of images of a fence panel. Unfortunately the Batch Processor will not work properly (output files are too dark) so I had to do all the work by hand, with a few judicious macros.

I attach here one of the files resized 200%. I stacked 13 files from a burst-mode hand-held group and exported the median and the mean. All were identical to this sample image! I even tried layering them with a difference blending mode (which was black). I suspect that my subject matter is not really suitable for this technique.

SampleCrop2X2.thumb.png.4a0f9947ad2e3b7df031aef0a4c45a7e.png

Windows 10, Affinity Photo 1.10.5 Designer 1.10.5 and Publisher 1.10.5 (mainly Photo), now ex-Adobe CC

CPU: AMD A6-3670. RAM: 16 GB DDR3 @ 666MHz, Graphics: 2047MB NVIDIA GeForce GT 630

Link to comment
Share on other sites

5 minutes ago, owenr said:

@dronecrasher, on 1st April you posted a zip containing 7 DNGs and a "TEST2.JPG". That JPEG is a definite success - it has an unambiguous increase in detail when compared to the processed DNGs, but I cannot get AP to produce anything like as good a result, so how did you do it? Was it done by other software? Is the date of 1st April significant?

 

Here are screenshots of one of the processed DNGs, then two Affinity results, and then your "TEST2.JPEG".

Click them to access the real images instead of forum previews.

 

dji_0131.thumb.png.d98fc96c892dcf614d8cb6747e5a810a.png

5aca526fb92df_meanauto-alignedstackresult.thumb.png.46fede458639092b3a7a0ac84821827f.png

5aca52a66cbbb_manualopacitystackmerged.thumb.png.ee9f0cde3e6160ec6d65935a163ed210.png

test2.thumb.png.73af2ffbf6fd761c18a5d279bef775a5.png

 

 

I adjusted the sharpness before I exported it into a JPEG. I've noticed afterwards that it is the wrong way to determine if super resolution really works since I am not comparing apples with apples and I should not make any adjustments 

 

 

Link to comment
Share on other sites

4 hours ago, dronecrasher said:

I think it works we just have not figured out how to do it. In the explanation https://petapixel.com/2015/02/21/a-practical-guide-to-creating-superresolution-photos-with-photoshop/ it states to not auto align while stacking otherwise it won't work. but how do I auto align the images/layers?

I read this petapixel article some time ago. It was what made me think "That's interesting, I must try it sometime." I did not actually try it out until @dronecrasher started this thread.

Alignment is indeed an interesting aspect. I have been re-sizing first, then stacking with alignment. I must try it without the auto-alignment.

I have also been trying out PhotoAcute. I have given it a run, but have not yet compared results.

John

Windows 10, Affinity Photo 1.10.5 Designer 1.10.5 and Publisher 1.10.5 (mainly Photo), now ex-Adobe CC

CPU: AMD A6-3670. RAM: 16 GB DDR3 @ 666MHz, Graphics: 2047MB NVIDIA GeForce GT 630

Link to comment
Share on other sites

1 minute ago, owenr said:

Please post your result before the sharpen was applied because my best result from your 7 DNGs cannot possibly be sharpened into an image like your TEST2.JPG.

Did you maybe use far more then 7 DNGs?

 

 

I only used 7 DNGs.

 

Here is my latest result and what I've done:

  1. imported 7 raw files to affinity
  2. develop persona - bottom right where it says basic I disabled enhance
  3. develop persona - bottom right where it says details I unchecked everything incl. noise reduction
  4. applied the same settings to all DNGs and developed them
  5. resized every single image by 200% and selected Resample Lanczos 3 (non-separable)
  6. exported as JPEG (highest quality)
  7. imported all 7 JPEG via stack method, auto align selected
  8. no adjustments made, export as final.jpeg

 

I attached four images (left DNG / right Final.jpeg) to show you the results. I also attached one of the 7 original raw files and the final.jpeg

 

DJI_0153.DNG

finalversionapril8.jpg.zip

Screen Shot 2018-04-08 at 2.04.02 PM.png

Screen Shot 2018-04-08 at 2.01.44 PM.png

Screen Shot 2018-04-08 at 2.00.30 PM.png

Screen Shot 2018-04-08 at 1.59.47 PM.png

Link to comment
Share on other sites

23 minutes ago, dronecrasher said:

5. resized every single image by 200% and selected Resample Lanczos 3 (non-separable)

The petapixel article article step 2. Resize to 200% Width/Height says:

Quote

Choose Image>Image Size…
· Set Width/Height to 200%
· Use the “Nearest Neighbor” resample algorithm. You can also use “Preserve Details” but I prefer “Nearest Neighbor” as it does not oversharpen
· Click OK

I am not sure but I believe you will not get good results if you use Lanczos 3 resampling.

All 3 1.10.8, & all 3 V2.4.1 Mac apps; 2020 iMac 27"; 3.8GHz i7, Radeon Pro 5700, 32GB RAM; macOS 10.15.7
Affinity Photo 
1.10.8; Affinity Designer 1.108; & all 3 V2 apps for iPad; 6th Generation iPad 32 GB; Apple Pencil; iPadOS 15.7

Link to comment
Share on other sites

1 hour ago, R C-R said:

The petapixel article article step 2. Resize to 200% Width/Height says:

I am not sure but I believe you will not get good results if you use Lanczos 3 resampling.

I am aware of them saying nearest neighbor but the lanczos 3 option supposedly sharpens the edges more aggressively and after trying it out (my previous post) I was satisfied with the result.

 

 

Link to comment
Share on other sites

2 hours ago, owenr said:

 

I've been unable to get good results in Affinity after using nearest neighbour to upscale 7 developed raw files, so I'll be trying the other resampling methods when I've time.

 

 

It's weird but I just tried out this method and it got me the best results so far.

 

Downside : my final JPEG file was 120mb and the megapixel went up to 192MP

 

Record a macro after opening a file of the series (View-> Studio -> Macro activates the side panel)

- Press the red record button in the macro panel

- Apply the changes you want to be made (adjustments + resize document)

- Press the "Add To Library" button in the macro panel (the one with the three squares and the plus)

   - Use the category "Macro" and give it an name you can later easily recognize what it does (e.g. "Super Resolution Preps")

 

Now go to File -> New Batch Job

- Hit the "Add" button and add all photos you want to be processed

- On the lower right side of the dialog under "Available Macros" select Macros in the drop down menu and chose the macro you created. Hit Apply.

- On the upper right side select the file type you want your photos to be saved as (but not the affinity format because you cannot add them to a stack later on)

- Hit OK and get a coffee ;)

 

After everything finished, go to File -> New Stack

- Add the files processed by the macro

- Choose the aligning options you want

- Hit OK and get another coffee

- In the "Layers" tab on the "Live Stack Group" choose the blending method you like (the symbol next to the check mark)

 

Screen Shot 2018-04-08 at 6.08.41 PM.png

Link to comment
Share on other sites

3 hours ago, dronecrasher said:

I am aware of them saying nearest neighbor but the lanczos 3 option supposedly sharpens the edges more aggressively and after trying it out (my previous post) I was satisfied with the result.

Hmmm. I replied to this around 20 minutes ago but the post seems to have vanished since then. :(

 

Briefly, what I said was that (AFAIK) lanczos resampling does not sharpen; in fact it applies a low frequency smoothing filter that reduces the aliasing effects of hard edged resampling methods like nearest neighbor. The point of upsampling using nearest neighbor resampling is to make sure each upsampled pixel in each photo is transformed into 4 identical pixels so that when the upsampled photos are aligned & averaged afterwards, the averaging process includes as much of the variation in each upsampled pixel group as possible, which is where the 'hidden' resolution comes from.

All 3 1.10.8, & all 3 V2.4.1 Mac apps; 2020 iMac 27"; 3.8GHz i7, Radeon Pro 5700, 32GB RAM; macOS 10.15.7
Affinity Photo 
1.10.8; Affinity Designer 1.108; & all 3 V2 apps for iPad; 6th Generation iPad 32 GB; Apple Pencil; iPadOS 15.7

Link to comment
Share on other sites

This topic has been too interesting. I find that I have been spending ages doing multiple upsizes (because the batch processor won't behave) and spending ages waiting for the stack to be built. I have also been using the trial version of PhotoAcute, which also takes ages to create an output file (complete with watermark).

When I have a spare week, I might come back to it. Meanwhile, good luck to all you who are persisting with SuperResolution!

John

Windows 10, Affinity Photo 1.10.5 Designer 1.10.5 and Publisher 1.10.5 (mainly Photo), now ex-Adobe CC

CPU: AMD A6-3670. RAM: 16 GB DDR3 @ 666MHz, Graphics: 2047MB NVIDIA GeForce GT 630

Link to comment
Share on other sites

  • 1 year later...

I came across this thread by chance, and tried PhotoAcute3 - now no longer developed I think. Anyways, with jpegs it didn't work too well, but with RAW files things went better.

Since it couldn't read GH5 RAW, I had to pretend it was a GH3, and the colours went all wrong. De-saturated and posted her for anyone else wondering if it's worth it.
PhotoAcute makes DNG files that it works with, BTW., so it's not really RAW files, but converted RAW files. 
Will now try Affinity, but I believe people have trouble with:

1) Stacking that tries to hide imperfections with blending etc. will for obvious reasons not work

2) Re-sizing needs to maximise variance, so closest neighbour enlargement is crucial

I'll report back any results.

GH5 with Olly 12-40, 1600 ISO 8 exposures hand-held with RAW and jpeg. RAW used in pictures below.

 

SuperResolutionRAWGH5w8images1600ISO.thumb.png.3f7d1f264093e92889452ec5edaa4b27.png

 

Edited by Amotilos
Forgot to mention picture data
Link to comment
Share on other sites

So this is the result of using Affinity Photo. A massive reduction in colour noise, and with it, a jump in resolution. So, for ISO 1600 files shot in the dark, yes, it's a worthwhile technique.

What I did:
1) Imported all the 8 RW2 files into AF
2) Just pressed "Develop"
3) Pressed EXPORT as TIFF 16 bit colour and changed size in export options box to 2x in both dimensions, and chose "nearest neighbour"
4) Chose IMPORT stack, just a normal stack, with align
4) EXPORTed a 97% jpeg in normal size as the final result.

Hope this may inspire someone. I'll be trying out ISO 100 on a sunny day with the lens stopped down to f8.0 next... :-)
 

2128573997_ScreenShot2019-08-20at20_17_49.thumb.png.1059e94bdd4c2ab79e4d593f668086cc.png

Link to comment
Share on other sites

  • 1 year later...

I haven't had any luck with the technique; other than noise reduction which is awesome!

I also can't understand the rational behind upscaling the images to 200% before importing them into the stack. What's the problem with upscaling them after the stack it created? It's dynamic anyways - isn't it?

Unless this is for the auto-alignment algorithm to have more precision in alignment (though to be honest I find the alignment very ineffective) - Any thoughts on that?

Link to comment
Share on other sites

Following @F_Kal's comment I did a little trial. I first followed the usual protocol using the same 13 images of a fence panel.I enlarged each 200% (with Lanczos), then stacked them with alignment and saved the median value, then cropped it.

1593900092_FencePanelMedianX2crop.jpg.12babbf071f999d975afe00062960d55.jpg

Secondly, I stacked the original images with alignment, then I enlarged 200% (with Lanczos) then cropped.

FencePanelMedianX1Crop.png.735e528d68935cccc7688bad866fb3f9.png

Both images are displayed at 100%. If anything, I find that the second one (stack then enlarge) has the sharper result (although neither are tack-sharp).

John

Windows 10, Affinity Photo 1.10.5 Designer 1.10.5 and Publisher 1.10.5 (mainly Photo), now ex-Adobe CC

CPU: AMD A6-3670. RAM: 16 GB DDR3 @ 666MHz, Graphics: 2047MB NVIDIA GeForce GT 630

Link to comment
Share on other sites

22 minutes ago, John Rostron said:

If anything, I find that the second one (stack then enlarge) has the sharper result

The second does (counterintuitively) look decidedly sharper to me.

Alfred spacer.png
Affinity Designer/Photo/Publisher 2 for Windows • Windows 10 Home/Pro
Affinity Designer/Photo/Publisher 2 for iPad • iPadOS 17.4.1 (iPad 7th gen)

Link to comment
Share on other sites

I’d say the second one looks like it has an unsharp mask and a bit of colour adjustment.

iMac 27" 2019 Somona 14.3.1, iMac 27" Affinity Designer, Photo & Publisher V1 & V2, Adobe, Inkscape, Vectorstyler, Blender, C4D, Sketchup + more... XP-Pen Artist-22E, - iPad Pro 12.9  
B| (Please refrain from licking the screen while using this forum)

Affinity Help - Affinity Desktop Tutorials - Feedback - FAQ - most asked questions

Link to comment
Share on other sites

interesting finds @John Rostron!

Could you elaborate a bit on the alignment method you used? was it auto-align or was it manually micro-tuning the individual images' perspective?

Also, how do the results compare to the original images? Any chance you could  share the crop of one of the original 13 images for us to see?

Thank you for helping out!

Link to comment
Share on other sites

3 hours ago, F_Kal said:

Could you elaborate a bit on the alignment method you used? was it auto-align or was it manually micro-tuning the individual images' perspective?

It was auto-aligning.

3 hours ago, F_Kal said:

Any chance you could  share the crop of one of the original 13 images for us to see?

I'll see what I can do.

John

Windows 10, Affinity Photo 1.10.5 Designer 1.10.5 and Publisher 1.10.5 (mainly Photo), now ex-Adobe CC

CPU: AMD A6-3670. RAM: 16 GB DDR3 @ 666MHz, Graphics: 2047MB NVIDIA GeForce GT 630

Link to comment
Share on other sites

thank you John! Interesting; I've found auto-alignment to be most of the time producing inaccurate alignments - even with images from a tripod (let alone handheld)

I also wonder whether this technique only yields results with extremely sharp lenses, where the sensor doesn't resolve all the detail procured by the lens. The kit and adapted lenses from the 80s that I use don't ever reach pixel-level sharpness

Link to comment
Share on other sites

16 hours ago, John Rostron said:

Any chance you could  share the crop of one of the original 13 images for us to see?

This is the middle image from a sequence of 13, resized 200% with Lanczos.

468843995_MiddleImagex2.png.341b90d7ddd69eceb8cf9a85dbe45ff4.png

John

Windows 10, Affinity Photo 1.10.5 Designer 1.10.5 and Publisher 1.10.5 (mainly Photo), now ex-Adobe CC

CPU: AMD A6-3670. RAM: 16 GB DDR3 @ 666MHz, Graphics: 2047MB NVIDIA GeForce GT 630

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
×
×
  • Create New...

Important Information

Terms of Use | Privacy Policy | Guidelines | 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.