SilverFast prescan images have artifacts that affect selection of black point and white point

[JokerNZ]

tldr; Has anyone using Silverfast software noticed edge ringing in prescan images that seem to affect automatic colour correction?

---

I've been using Silverfast 8 and now 9 for a while. Currently I'm scanning thousands of family slides.

Just recently I noticed a problem with prescans.

I use an Epson V800 flatbed scanner. I do my prescans at high res x4. I am prescanning a set of slides in a slide holder.

Example of slide holder layout

If I have some relatively high contrast edges, a prescan produces edge artifacts. Below is an example.

On the left is a prescan image. You can see the darker ringing where the white tree meets the darker BG. You can also see a sampler on the left of the image, labelled with a B. This is the point where the Silverfast automatic blackpoint picked. It has clearly picked a position with the BLACK ringing.

The right hand image is the slide scanned at full res, full colour depth, everything. Just to demonstrate that the ringing is not in the original slide.

Left: Prescan image at x4, showing ringing around bright to dark areas. RIGHT: full res scan to show the problem is not in the original

This would not really be a problem, except that the automatic colour correction tools seem to be using the blown up prescan, with the artifacts, to get blacks and whites. This will produce incorrect corrections, because the artifacts are from filtering, and blowing the image up.

If the automatic tools were using the original prescan image size, or a blown up version using a filter that doesn't produce edge artifacts, then I expect it would be fine.

I work in VFX, so I'm familiar with various filters and the kinds of artifacts they can produce. And what I'm seeing in the prescans are artifacts for filters like sinc, lanczos and similar.

In order to not get those artifacts it needs a filter like cubic, or bicubic I expect.

I have been having a conversation with a Silverfast support person, back and forth.

And he insists that Silverfast uses a bicubic filter for enlarging the prescans. That may be true for enlarging them to view on a monitor, inside Silverfast.

But I cannot see that as true for whatever enlargement is being done when sampling is taking place.

I don't know if I am missing something fundamental. Or if we are communicating badly.

Its possible that English isn't his first language (since its a German company) so I am trying to ask clearer questions, without being insulting with too simple language... :-/ Tech support is hard enough without subtle language differences causing confusion.

In any case, if anyone has any insight into this, I'd be very grateful.

Cheers

 

[koraks]

I don't use with VFX, but what I see in those prescans looks to me like applying an overly aggressive and coarse unsharp mask.

However, have you considered ignoring the whole thing and just profiling your scanner with an IT8 target? This should get your colors close to reality from the get go. At that point it's a matter of applying final color and/or contrast adjustments to taste, for instance to correct for the massive fading in the example slide you show.

 

[Radost]

tldr; Has anyone using Silverfast software noticed edge ringing in prescan images that seem to affect automatic colour correction?

---

I've been using Silverfast 8 and now 9 for a while. Currently I'm scanning thousands of family slides.

Just recently I noticed a problem with prescans.

I use an Epson V800 flatbed scanner. I do my prescans at high res x4. I am prescanning a set of slides in a slide holder.

View attachment 376787
Example of slide holder layout

If I have some relatively high contrast edges, a prescan produces edge artifacts. Below is an example.

On the left is a prescan image. You can see the darker ringing where the white tree meets the darker BG. You can also see a sampler on the left of the image, labelled with a B. This is the point where the Silverfast automatic blackpoint picked. It has clearly picked a position with the BLACK ringing.

The right hand image is the slide scanned at full res, full colour depth, everything. Just to demonstrate that the ringing is not in the original slide.

View attachment 376788
Left: Prescan image at x4, showing ringing around bright to dark areas. RIGHT: full res scan to show the problem is not in the original

This would not really be a problem, except that the automatic colour correction tools seem to be using the blown up prescan, with the artifacts, to get blacks and whites. This will produce incorrect corrections, because the artifacts are from filtering, and blowing the image up.

If the automatic tools were using the original prescan image size, or a blown up version using a filter that doesn't produce edge artifacts, then I expect it would be fine.

I work in VFX, so I'm familiar with various filters and the kinds of artifacts they can produce. And what I'm seeing in the prescans are artifacts for filters like sinc, lanczos and similar.

In order to not get those artifacts it needs a filter like cubic, or bicubic I expect.

I have been having a conversation with a Silverfast support person, back and forth.

And he insists that Silverfast uses a bicubic filter for enlarging the prescans. That may be true for enlarging them to view on a monitor, inside Silverfast.

But I cannot see that as true for whatever enlargement is being done when sampling is taking place.

I don't know if I am missing something fundamental. Or if we are communicating badly.

Its possible that English isn't his first language (since its a German company) so I am trying to ask clearer questions, without being insulting with too simple language... :-/ Tech support is hard enough without subtle language differences causing confusion.

In any case, if anyone has any insight into this, I'd be very grateful.

Cheers

Are you making adjustment before scan? I recommend upgrading your version to get RAW scan.

 

[Radost]

I don't use with VFX, but what I see in those prescans looks to me like applying an overly aggressive and coarse unsharp mask.

However, have you considered ignoring the whole thing and just profiling your scanner with an IT8 target? This should get your colors close to reality from the get go. At that point it's a matter of applying final color and/or contrast adjustments to taste, for instance to correct for the massive fading in the example slide you show.

As far as I know IT8 only works well only for positives.

 

[koraks]

As far as I know IT8 only works well only for positives.

Yup. Since that's what he's scanning, I figured it's an appropriate suggestion.

With negatives, I prefer to use Epson scan, turn off all auto-corrections and scan at 16b/channel and then correct manually in post.

 

[JokerNZ]

I don't use with VFX, but what I see in those prescans looks to me like applying an overly aggressive and coarse unsharp mask.

However, have you considered ignoring the whole thing and just profiling your scanner with an IT8 target? This should get your colors close to reality from the get go. At that point it's a matter of applying final color and/or contrast adjustments to taste, for instance to correct for the massive fading in the example slide you show.

No there's no sharpening of any kind being applied. I specifically turned it off for this example. Besides, any image sharpening should be happening only to the full size scan.

I have the iT8 profile charts for transparencies and and have calibrated using them.

When the slides in question have lasted well (they are from the 1940s to the mid 1980s) then they come up very nicely.

I am quite happy to use the auto correct functions for slides that have faded. Then judge whether to use the autocorrect, or adjust what the autocorrect has done.

I do a quick balance to the prescans, and always keep the low and top ends wide. So my scans tend to look a bit lifted in the blacks, and lower contrast.

I deal with that in Lightroom.

The slide example is an example that shows the problem very clearly. I don't really expect to recover that image. I think that was how it was shot / developed.

 

[JokerNZ]

Are you making adjustment before scan? I recommend upgrading your version to get RAW scan.

I have considered it, but frankly for my purposes, its not worth the money. This is for my family, so I'm scanning to a high level, but probably not a pro photographer level. Especially since I'm using a Epson V800. If I find any really good slides, I'll get them scanned on a drum scanner or similar.

I am scanning and saving at 16bit, reasonably high resolution, and keeping the histogram wide to encompass the data I can see in the histograms.

Then correcting in Lightroom.

Lately I am also becoming reluctant to give Silverfast any more money. (Lack of an actual manual for Silverfast 8 or 9, and this weirdness that they seem to be weirdly not recognizing. Or at least coming back to me with anything sensible.)

So I am also trying to keep my images platform neutral, since I don't want them trapped in Silverfast.

Sure, if I get rid of Lightroom, I may well lose the Lightroom corrections. But the original images will still be accessible.

 

[koraks]

The slide example is an example that shows the problem very clearly.

I'll take your word for it; I mean, I'm not sure what's to be expected from this image. I do see the strong edge effect you appear to be referring to, but I can't tell if it's the cause of the auto balancing feature failing on this scan. Maybe the Vuescan people could at least verify this for you.
Either way, I'd just manually balance a scan like this (or any scan, really).

 

[brbo]

@JokerNZ, if LaserSoft can't comprehend the nature of the problem (it would be totally in line with that company) your only solution is to not use SilverFast auto-correction (and why would you use that in the first place if your workflow does not start and end with SilverFast?).

I haven't really used later versions of SilverFast (8 and 9) much, but I still use older versions of SF with certain scanners when OEM software does not work anymore or is inferior even to SilverFast. At least in v6.x you could still scan in full DRange mode (no auto adjusting). If you also scan your IT8 target in that mode and then make a profile from that you will have something that is close to what original slides look like after you apply this profile to your full DRange scans.

 

[jeffreyg]

I believe you should have the scanner attachment for scanning on the glass without the holder. Try scanning with that and not the holder to see if that makes a difference. I used that with some forty year old slides I happened to run across and the scans were very accurate even the color.