darktable icon indicating copy to clipboard operation
darktable copied to clipboard

Snapshot preview esp when zoomed in does not match preview of the same image used to make the snapshot

Open todd-prior opened this issue 1 year ago • 8 comments

Describe the bug

So if you take a snapshot of an image and then you activate that snapshot you notice a less contrasted saturated version that does not match the image .. zooming to 100 percent makes it clear. In one case its the snapshot that looks like this in the second case it was the current image preview. I will include both the xmp files. If you enable HQR then the snapshot and preview match as expected. This could potentially impact comparing different images that could be altered without you knowing..

Steps to reproduce

Open an image in darkroom. Take a snapshot Enable that snapshot Zoom in THe snapshots and image preview should match but do not unless HQR is enabled This happens with or without OPEN CL

Expected behavior

Snapshot and preview should match

Logfile | Screenshot | Screencast

No response

Commit

I saw this first in +1237 and still now in tonights master

Where did you obtain darktable from?

self compiled

darktable version

4.7.0+1334~gf3245d1a5b

What OS are you using?

Windows

What is the version of your OS?

Windows 11 Pro

Describe your system?

!2th Gen intel with 32BG and 3060TI video card

Are you using OpenCL GPU in darktable?

Yes

If yes, what is the GPU card and driver?

Nvidia 3060Ti

Please provide additional context if applicable. You can attach files too, but might need to rename to .txt or .zip

https://github.com/darktable-org/darktable/assets/44409419/ce225a9d-1965-45dd-ab97-5a175b7e63aa

https://github.com/darktable-org/darktable/assets/44409419/8d697b71-f9c2-4265-bbea-cf4f599f8ac8

PK320007_02.DNG.zip

Images

https://discuss.pixls.us/t/differences-in-the-output-sharpness-of-jpg-when-processed-with-dt-or-lr/43724/6

https://discuss.pixls.us/t/hoodoo-hike-have-fun-i-did/43638

todd-prior avatar Jun 03 '24 00:06 todd-prior

@jenshannoschwalm Thanks for checking... I will confirm when I get home tonight

todd-prior avatar Jun 06 '24 17:06 todd-prior

I removed that comment as not correct. Would help a lot if you could find some "pattern" like "these modules are problemetic". If we are lucky and you can provide some good information we might be able to get this sorted out for 4.8

jenshannoschwalm avatar Jun 06 '24 17:06 jenshannoschwalm

Okay I will. I am out until about 8 or so EST tonight but I will poke around a bit when I get home and see... It doesn't happen when the new HQR preview is enabled but it does when it is not toggled on. I don't know if that is any clue to what might be happening or not with respect to pipeline calculations..... I will see if I can find out what it is and reply ASAP....

todd-prior avatar Jun 06 '24 18:06 todd-prior

Again, haze removal seems to be the bad boy. Also that algo is suspicious.

jenshannoschwalm avatar Jun 06 '24 19:06 jenshannoschwalm

On the first image the hoodo's image ...I walked throught the xmp.... three modules seemed to introduce the effect or at least if I disabled the module and took the snapshot I would not see the difference... it was dehaze, chromatic abberations, and denoise profiled...... not sure if that helps.... no other module disabled removed the difference that could be noted...

todd-prior avatar Jun 07 '24 03:06 todd-prior

Second example only dehaze seemed responsible...not the others....

todd-prior avatar Jun 07 '24 03:06 todd-prior

The haze removal module has a known bug in computing the base value for the detected haze, which then impacts the strength of the removal. There's actually a fix already in the code, but I didn't enable it at the time because it requires a version bump and I didn't have time to figure that out. (The bug is that the quickselect implementation doesn't find the 95th-percentile haziest pixel, it finds some random one....)

ralfbrown avatar Jun 24 '24 07:06 ralfbrown

There is at least one more issue with module requiring full image area once (done I preview pipe) for "calibration"

jenshannoschwalm avatar Jun 24 '24 15:06 jenshannoschwalm

This issue has been marked as stale due to inactivity for the last 60 days. It will be automatically closed in 300 days if no update occurs. Please check if the master branch has fixed it and report again or close the issue.

github-actions[bot] avatar Aug 24 '24 00:08 github-actions[bot]

So this file https://discuss.pixls.us/t/a-hummingbird-play-raw/45085 This xmp 619A6812_01.CR3.zip

https://github.com/user-attachments/assets/e76a268a-fdfb-445d-9028-43cd7f8b6e08

Looking at the head..the snap shot is the edited raw...initial paired with itself and detail is equivalent... advance to the jpg and there is a visible softening of the snapshot preview... detail returns when you go back to the raw in the filmstrip...

todd-prior avatar Aug 30 '24 04:08 todd-prior

This issue has been marked as stale due to inactivity for the last 60 days. It will be automatically closed in 300 days if no update occurs. Please check if the master branch has fixed it and report again or close the issue.

github-actions[bot] avatar Oct 30 '24 00:10 github-actions[bot]

@todd-prior Does this problem still occur in 5.0?

ralfbrown avatar Dec 28 '24 10:12 ralfbrown

I'll check and update

Sent from Outlook for Android

From: Ralf Brown @.> Sent: Saturday, December 28, 2024 5:48:52 AM To: darktable-org/darktable @.> Cc: Prior,Todd @.>; Mention @.> Subject: Re: [darktable-org/darktable] Snapshot preview esp when zoomed in does not match preview of the same image used to make the snapshot (Issue #16925)  

Caution: External email.

@todd-prior Does this problem still occur in 5.0? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>

todd-prior avatar Dec 28 '24 14:12 todd-prior

This issue has been marked as stale due to inactivity for the last 60 days. It will be automatically closed in 300 days if no update occurs. Please check if the master branch has fixed it and report again or close the issue.

github-actions[bot] avatar Feb 28 '25 00:02 github-actions[bot]