Weird blurry patterns when exporting

[fritzunruh]

Hey there,

 

today I took some TIF-files to my photolab of choice to get large scale prints of scanned and retouched film.

 

Going over the pictures, the lab-guy noticed an odd blurry situation with the grain on one of them. He was certain it had to do something with a faulty scanner on my end, which I was sceptical about, back home I checked and the original file doesnt show any of said weirdness. I'm used to develop all my pictures inside of Adobe Bridge with Camera RAW, where you could not see the issue at all.

 

I also tried to export the same picture from Lightroom, again it does only appear on the exported uncompressed 16-bit TIF as well as on a PNG I tried for troubleshooting.

 

I'm fairly new to working with scans and prints as well, so I might miss something very obvious, but I was under the impression that TIF is a lossless format that should not compress or alter what I see inside of CRAW or LR at all unless specified?

 

Thanks a lot!

 

Full Image:

 

 

Grain Pattern inside of LR (thats what it looks on the original scan data as well)

 

 

Grain Pattern when exported to TIF or PNG

 

[Ed_Ingold]

Look like grain reduction artifacts. Should be avoided when scanning B&W. ICE, which uses infrared, fails completly with B&W because silver is opaque to IR and can't be distinguished from dust.

[fritzunruh]

Ok, but the scan wasn't the issue here? That looks fine! Its a color film, too. I suspected my digital noise reduction might be the source of the problem, but that doesnt explain why only the export is distorted, does it? It looks like a badly compressed jpeg to me, not like a 35 MB TIF/PNG file?

[john_wheeler6]

Hi fritzunruh

To diagnose the problem the original files would be needed for both. It was not clear that the full image you provided was the full original image or a smaller size and/or i bit when started at 16 bit. Are the other images screen shots of enlarged portions?

The issue is that screenshots at non 100% viewing having gone through interpolation from the actual data and as well even at 100% typically go through color mangement so you are not looking at the original data.

Also, was the image resized by the printer of did you resize yourself before sending to your printer. Resizing also goes through an interpolation and depending on the rendering algorithm can introduce anomalies.

So first step is to have the exact files (or crops with no enlargements/reductions) from proper comparisons at 100% magnification by forum members.

I have seen such issues yet the root cause is difficult without the exact original data bits. Just a suggestion

John Wheeler

[Ed_Ingold]

Resizing and resampling involves interpolation of blocks of pixels, similar to what occurs in digital noise reduction. I see what I call "worm tracks" in the second image, ie artifacts near the pixel level. If you can't see it in a print at a "normal" viewing distance, I wouldn't worry about it.

[fritzunruh]

Thanks, yes, the last two screenshots are enlarged portions of the image, once inside of LR with no artifacts, second one shows the artifacts after exporting.

 

At 100% size, it looks exactly like it does in my screenshots. There was no resizing of the image done at all when exporting, still I seem to get compression artifacts when Im in fact not compressing anything while exporting to 16bit TIF without compression or resizing, am I?

 

If I can see it on a print at normal viewing distance greatly depends on the size that I'm printing it, I'd rather just work with a flawless image without artifacts than having to hope for the best while spending lots of money on a high quality print.

[fritzunruh]

This is a screenshot of an enlarged portion of the picture inside the LR module - There are no artifacts at all

 

 

This is the original file without any edits, just converted from tif to png to be able to upload here - no artifacts at all of course

 

 

This is the exported final result as a jpeg, just because I cant upload a tif here - but the issue with a tif is identical. Of course the jpeg shows some additional signs of compression, but the blurred areas of smudged grain look just like that on a uncompressed file. To see it, you have to zoom in a little, but not to a pixel level.

 

 

It drives me nuts.

[Ed_Ingold]

The original and copy appear to be at a different scale. Pixels in the "distorted" version are roughly twice the size of those in the original. This suggests resizing and/or resampling. You need to examine your work flow carefully to see where this might be occurring.

 

A RAW scan is not the equivalent of a RAW digital image from a camera. It is merely a TIFF (or other) file with little or no processing by the scanning software.

 

A screen shot involves resampling on at least planes - image to screen and screen to image. Open the image in an image processor (e.g., Photoshop or Lightroom)) and export a crop of that image taken at a fixed pixel size, such as 900x900. You can also view the image at 1:1 pixel size. Depending on the resolution of your monitor, you may need to magnify the image 3x or 4x to see the actual pixels in the image.

[fritzunruh]

You're right, the pixels of the export appear to be bigger. But how can that be the case when I'm exporting without any resizing of the image? The dimensions are the same too, still it seems it looses half the resolution!

 

I tried to make a fixed pixel sized comparison, but realized that when I import the source image into photoshop I can't get around the "development" done by the camera RAW module, which leads to the compression - even though I didnt even export the image!

 

I tried to work around that by screengrabbing a fixed rectangle of a evenly magnified osx preview as seen below.

 

Original

 

 

After

 

[Ed_Ingold]

How was the image scanned? Into what format, size (pixels), etc?

[fritzunruh]

scanned using silverfast into a tif with 2903 × 1943 pixels, not sure about the model of the scanner but a rather advanced type, no sub 1k $ consumer model afaik

 

I get that these specs help generally, but in this case it seems that the issue lies somewhere in the processing of the data rather than the capturing? The source file looks good, just when put into the Adobe Camera RAW processing system it messes with the compression somehow and leads to a noticeable drop in quality.

[fritzunruh]

To add something I just discovered:

 

I tried to open the original source image from my scan in photoshop and reset all development settings upon import (Camera RAW pops up and wants settings, I just reversed everything and opened the image)

 

The issue comes up even before exporting the image, as soon as the source material is "touched" by got "out of" the Camera RAW development module the image muds up, even inside of photoshop!

[digitaldog]

Visible banding can be caused in two vastly different areas: The data itself OR the display path (perhaps both).

If the image is high bit (16-bit per color even though it's rarely that encoding), the banding seen is likely instead in the display path. IOW, it's not in your data. That may explain why you don't see it in Develop at 1:1, the proper way to view the data in the first place. You can't evaluate banding at less than 1:1 or 100% due to a zoomed out preview uses differing subsampling to show you the image.

The banding in the display path can be due to it not being totally high bit either. In a high bit display path, ALL areas must be high bit: The video card, the display panel, the software and the OS.

[fritzunruh]

Visible banding can be caused in two vastly different areas: The data itself OR the display path (perhaps both).

If the image is high bit (16-bit per color even though it's rarely that encoding), the banding seen is likely instead in the display path. IOW, it's not in your data. That may explain why you don't see it in Develop at 1:1, the proper way to view the data in the first place. You can't evaluate banding at less than 1:1 or 100% due to a zoomed out preview uses differing subsampling to show you the image.

The banding in the display path can be due to it not being totally high bit either. In a high bit display path, ALL areas must be high bit: The video card, the display panel, the software and the OS.

 

Thanks! I realize a banding issue can occur when an image is zoomed out and its physics with the display, but in my case the "compression look" is seen on every zoom-level as you can see above? If the issue just lays in my hardware chain another display would not have the issue, right? I double checked with an eizo color edge, same problem persists. And my own setup being a recent and maxed out Macbook Pro in combination with up-to-date Adobe software should be able to process an image without such a noticeable drop in quality, shouldn't it?

[digitaldog]

The Eizo should be a high bit panel but the video card?

If you don't see any banding, and it's high bit data, you're good to go. And sure, it's possible there is banding in that data, certainly if you see it on a high bit system or the data is indeed high bit (where it came from is the $64K question).

It's possible it came about prior to being 'converted' to high bit.

Photoshop and newer Mac OS are indeed high bit capable but you'll need to check the video card. Likely it is. You can get this info from System Report (About this Mac), Grahics>Display as well as looking in Photoshop's General Preferences > Performances, Advanced Graphics Processor Settings (30-bit display).

[kmac]

I used to get banding similar to this. The only way I could fix it was to set multi-sample at 2 or 3 before scanning. Now, when I zoomed your image at different sizes, the banding kept changing appearance but always retained the same parallel evenness. Such evenly space bands represent the tracking of the scanner's sensor, they're seen mostly in light areas of the images but you'll notice they extend down to the dark part of the sea, they go right through the image, you just don't see that because they get fainter and fainter. You may have to experiment with different resolutions and multi-sampling till the bands are negligible.

 

[fritzunruh]

Thanks a lot for taking the time, it makes total sense that the bending comes from the scanners sensor, what I cant get my head around is the fact that its not there in the original scanned file but only appears after I export it from lightroom / photoshop / camera raw whatever. And its not the editing either, since it looks absolutely fine in the preview of those programs, no matter the zoom setting.

 

I wouldnt mind a slight shift of brightness that you made visible very well, I do mind the muddy and blurred grain pattern, which just is not in the original file!

[fritzunruh]

If anyone is eager to try, here is the original tif-file I got from the scanner: 0828.tif

 

When I load it into LR or open it with Camera RAW all is well, as soon as I export it, the grain is blurred out - even with no edits at all. Could it really be my system or does anybody have that problem?

[fritzunruh]

The Eizo should be a high bit panel but the video card?

If you don't see any banding, and it's high bit data, you're good to go. And sure, it's possible there is banding in that data, certainly if you see it on a high bit system or the data is indeed high bit (where it came from is the $64K question).

It's possible it came about prior to being 'converted' to high bit.

Photoshop and newer Mac OS are indeed high bit capable but you'll need to check the video card. Likely it is. You can get this info from System Report (About this Mac), Grahics>Display as well as looking in Photoshop's General Preferences > Performances, Advanced Graphics Processor Settings (30-bit display).

 

Yes, its a 30bit display with an 4GB Radeon Pro 560 GPU.

 

I see banding on every screen, setup and level of zoom, just after I exported the image - not prior.

[john_wheeler6]

Thanks! I realize a banding issue can occur when an image is zoomed out and its physics with the display, but in my case the "compression look" is seen on every zoom-level as you can see above? If the issue just lays in my hardware chain another display would not have the issue, right? I double checked with an eizo color edge, same problem persists. And my own setup being a recent and maxed out Macbook Pro in combination with up-to-date Adobe software should be able to process an image without such a noticeable drop in quality, shouldn't it?

 

After using your original image I could not recreate you issue. Comparing your two provided images, I did fine a 75 pixel x 75 pixel grid of pattern that coincided with the appearance of blur. Note that is a preceived blur yet I believe is a subtle change in luminosity between the sharp and duller areas.

 

You can see the pattern that I could extract in the image below.

 

This is definitely in the processing pipeline somewhere and apparently not in the magnification or slight resizing.

 

I suggest several things

1) Turn off GPU acceleration and see if it goes away.

2) Make sure you have the most recent versions of you OS, PS, LR, ACR, and GPU firmware (could be bugs discovered before)

3) Let forum members know which OS, PS, LR, ACR, and GPU you are using for further help.

 

[fritzunruh]

After using your original image I could not recreate you issue. Comparing your two provided images, I did fine a 75 pixel x 75 pixel grid of pattern that coincided with the appearance of blur. Note that is a preceived blur yet I believe is a subtle change in luminosity between the sharp and duller areas.

 

You can see the pattern that I could extract in the image below.

 

This is definitely in the processing pipeline somewhere and apparently not in the magnification or slight resizing.

 

I suggest several things

1) Turn off GPU acceleration and see if it goes away.

2) Make sure you have the most recent versions of you OS, PS, LR, ACR, and GPU firmware (could be bugs discovered before)

3) Let forum members know which OS, PS, LR, ACR, and GPU you are using for further help.

 

 

Thanks John.

 

The graphic acceleration hint brought another tiny piece to the puzzle. As soon as I turn it off, the issue can be seen inside of the preview LR as well. When I turn the acceleration back on, the issue is gone in the preview (as seen in my first posts screenshots). Unfortunately turning it off or on doesn’t help the exported result.

 

Apparently the hardware acceleration is needed to prevent what is perceived as blurry, but is not used while rendering the image to export - is that just my system?

 

Current versions:

OS: MacOS Catalina 10.15.3 (I'll update to 10.15.4 after this reply)

GPU: Radeon Pro 560 (A0.927, drivers linked to OS)

PS: 21.1.2 (up to date)

LR: 3.2.1 (up to date)

ACR: can' tell the version number, but up to date

[Ed_Ingold]

Adobe Camera Raw is a rather blunt tool for rendering images. I was curious why Adobe Bridge would treat TIFF images from a scanner as though they were RAW. Bridge has not been my image catalog since Lightroom was invented, and I never installed in on my Mac. Until you establish another default image editing program, Bridge will use Camera Raw for everything. RAW images from a digial camera will still resort to Camera RAW, even if you open them in Photoshop.

 

Camera RAW defaults are found in an unobtrusive line near the bottom of the window. I noticed immediately that my 16 bit TIFF files were listed as 8 bit. Clicking on that line opens the size and format defaults for ACR, which could be the reason the pixel size is doubled. One of the options is resizing. There seem to be discrepencies in color space too. (Bridge goes back into the dust bin for me.) Of course you can keep the same number of pixels and change the size or resolution (in PPI) without resampling. Photoshop and Lightroom don't care - the display is unchanged. However not all display programs are as forgiving, and size rather than pixels matters when setting up a print or print preview.

 

The vertical banding shown above appears to be due to aliasing or Moire patterns from the resampling process. A less obvious source of banding occurs when images of a low gradient in medium tones (e.g., blue or grey sky) are displayed on a monitor with 8 bit color depth. There are only 256 possible tone densities in 8-bit resolution, so the dividing lines become visible. You see them on the monitor, but they're not necessarily in the image.

[fritzunruh]

Adobe Camera Raw is a rather blunt tool for rendering images. I was curious why Adobe Bridge would treat TIFF images from a scanner as though they were RAW. Bridge has not been my image catalog since Lightroom was invented, and I never installed in on my Mac. Until you establish another default image editing program, Bridge will use Camera Raw for everything. RAW images from a digial camera will still resort to Camera RAW, even if you open them in Photoshop.

 

Camera RAW defaults are found in an unobtrusive line near the bottom of the window. I noticed immediately that my 16 bit TIFF files were listed as 8 bit. Clicking on that line opens the size and format defaults for ACR, which could be the reason the pixel size is doubled. One of the options is resizing. There seem to be discrepencies in color space too. (Bridge goes back into the dust bin for me.) Of course you can keep the same number of pixels and change the size or resolution (in PPI) without resampling. Photoshop and Lightroom don't care - the display is unchanged. However not all display programs are as forgiving, and size rather than pixels matters when setting up a print or print preview.

.

 

Noted, but ACR was just one of the applications resulting in the issue? PS and LR do the exact same thing.

 

The vertical banding shown above appears to be due to aliasing or Moire patterns from the resampling process. A less obvious source of banding occurs when images of a low gradient in medium tones (e.g., blue or grey sky) are displayed on a monitor with 8 bit color depth. There are only 256 possible tone densities in 8-bit resolution, so the dividing lines become visible. You see them on the monitor, but they're not necessarily in the image.

 

But they are in the image, you see them no matter the level of magnification?

[Ed_Ingold]

But they are in the image, you see them no matter the level of magnification?

Of course. I was trying to describe a different kind of banding (posterization) which is the result of poor tonal resolution in a monitor.

 

Vertical banding in a scan can be caused by dust on the sensor or optics near the sensor. However the bands tend to be irregular and randomly spaced. These are highly regular, and almost certainly due to beat patterns in sampling frequencies.

 

Horizontal banding is usually sharply defined and closely spaced. It is caused by errors in indexing the film or scan head, or variations between parallel sets of sensor arrays. They can be mitigated by selecting a finer scan that uses only one of the linear sensor arrays, or (e.g., printers) shorter indexing intervals.

Edited April 16, 2020 by Ed_Ingold

[john_wheeler6]

To help narrow down the issue a question and a suggestion

 

You said the version of LR was 3.2.1 I am confused as that would be Lightroom 3. Are you using the most recent version of Lightroom Classic or something else?

 

Though I am skeptical that it is related, you incoming image is in Adobe RGB color space and exporting you change to sRGB color space. This should not cause a problem yet who knows.

 

Also, there is PS and LR vs exporting vs using Save and Save As

 

To narrow down the issue I suggest a couple steps.

 

- In Photohosp, go to Preferences and turn off all GPU acceleration

- Also make sure that the Preferences are set to go directly into Photoshop and not through ACR

- Bring your original image into Photoshop and verify at 100% magnification (or 200 or 400%) that there is not fuzzy areas

- Note - make sure you keep the file in Adobe RGB and don't change to sRGB

- Use the File > Save As command and save as a PNG file (so no lossy compression)

- Bring that PNG file back into Photoshop and verify that the is no fuzzy areas

 

I will assume that it will come back just as the original with no fuzzy areas yet if wrong, just report back the problem and don't follow steps below

 

If you are at this point you have a baseline where you save the file and did not have a change. Each step below should be taken individually and not in combination by making the change in the steps above to test it out

 

1) Do the above yet first make sure that GPU acceleration is turned on in Photoshop

2) Do the above steps with the only change use the Save for Web legacy function (do not change to sRGB and leave that unchecked) (do not change image size)

3) Do the above steps with the only change being using the Export function in PS (with and without changing to sRGB)

 

4,5) Add GPU acceleration with #2 and #3

 

6-9) Change Preferences to go through ACR before entering into Photoshop and do the prior combinations.

 

I think one of the above combinations finally ends up with the problem occurring the way you said in Photoshop and this should help isolate which area is the root cause.

 

It is just the divide and conquer approach.

 

You could use using LR or not as a variable and which format you save in as a variable yet not sure that would be necessary.

 

Hope this helps track it down.