Pages: [1] 2   Go Down

Author Topic: What's wrong with SilverFast?  (Read 22008 times)

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
What's wrong with SilverFast?
« on: September 25, 2005, 12:08:02 am »

Whilst this post is a continuation of some issues raised in another thread, I think it probably deserves its own topic.

My first attempts scanning negatives using my Scan Elite 5400ll, put Dimage Scan first, Vuescan second and SilverFast last with regard to ease of use and color accuracy, within a reasonable time frame, of course.

I've since downloaded a more recent version of Vuescan which seems to have addressed a number of issues relating specifically to the 5400ll, and the order has now changed. Vuescan is by far the best, Dimage Scan second for an automatic scan with minimal adjustments, and SilverFast last again.

SilverFast's behaviour with negatives is a real puzzle. The prescans are not even in the ball park, not even near the ball park; in a completely different city in fact. There is clearly something seriously wrong here. The film type selections in Negafix seem to have no bearing on reality. They are completely inaccurate.

Reading Ian Lyons tutorial on Negafix, he mentions that the prescans can initially look way off because the crop lines might include the black edges surrounding the frame. Move the crop lines inwards and the color changes. Not so with my setup. Makes no difference to the over all color wherever the crop lines are. I've got no idea if this is another symptom of something seriously wrong, or if it's an improvement in the software to remove yet another pitfall that can confuse the amateur.

Of course, the big features of SilverFast are the numerous ways you can change color and color casts, and Negafix has added yet another range of color and hue adjustments. Given enough time, however bad the prescan may look, one can eventually knock it into shape. However, there's a major problem here, even if you do have the time. How do you really know when the colors are accurate? We can't tell by eyeballing a negative on a light-table.

On the issue of a scan into the ProPhoto color space looking redder than the same scan (with same adjustments) scanned into a smaller gamut color space such as Adobe RGB, this is indeed true with Silverfast and negatives, on my system.

Perhaps this is another clue as to incorrect settings somewhere. It's not just a case of the ARGB scan being less red. The entire color range is less saturated, including the appearance of the prescan, indicating that the 'numbers' have not changed with change of 'profile to embed' in the SilverFast CMS section. I'm getting an ARGB image with ProPhoto numbers. To get it looking right, as I initially edited it, I have to assign the ProPhoto color space in PS.

Nevertheless, whether I'm scanning slides or negatives, the prescan always matches the scanned result very closely, when opened in PS with whatever embedded profile I've specified in CMS.

I sure hope some of you gurus and SilverFast fans can shed some light on these issues. It's got me beat.
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
What's wrong with SilverFast?
« Reply #1 on: September 25, 2005, 12:44:42 am »

Ray, we covered alot of this territory in another thread that seems to have fallen victim to the great crash.

I am using the Minolta 5400 first model, you the second. I don't know how much of a difference that makes, but it is a difference.

I have worked extensively with Minolta's software, some with Vuescan (latest version) and by now quite extensively with Silverfast. My ranking is the exact reverse of yours.

I definitely find that I am getting the most colour-realistic scans from Silverfast, apart from this most recent problem with the red skin tones in ProPhoto colour space. The film preset I am supposed to use for my negative film (a retail branded version of Fuji Superia ASA 100), combined with a bit of tweaking in Negafix produces very credible prescans.

For me there are three tell-tale signs of good colour in a prescan: the grey balance for tones that I know are supposed to be grey, the hue of the sky and the colours of objects that I remember very well. Between one or more of these three clues, I have found it quite manageable to lock into believable colours at the scanning stage, usually with just a couple of grey "pipette" placers on the image.

For my set-up, Ian Lyons' admonitions to insure that the scan frame is within the picture area - and that there is no blank space between the film and the carrier frame are correct. Invariably, when a prescan comes out looking ridiculous it is  because either of these rules is broken.

Because ARGB is a smaller colour space than ProPhoto it stands to reason that images could be overall less saturated in the former than the latter, but not necessarily - depends on the image.

Jonathan has outlined an experiment he suggested I conduct for analyzing the ProPhoto reddening issue and to send him the results. That is for tomorrow - I shall do it. The feedback should be very informative.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
What's wrong with SilverFast?
« Reply #2 on: September 25, 2005, 09:54:09 pm »

Quote
Quote
I am using the Minolta 5400 first model, you the second. I don't know how much of a difference that makes, but it is a difference.

True. There are differences. One which is commented on quite often is the fact that the 5400ll uses a white LED light source whereas the earlier model uses a cold cathode fluorescent tube.

Quote
I have worked extensively with Minolta's software, some with Vuescan (latest version) and by now quite extensively with Silverfast. My ranking is the exact reverse of yours.

Of course my ranking only relates to scanning negatives. The situation with positives is quite different. Silverfast on the whole is preferred, but Vuescan can sometimes throw up a result that's spot on without hassle when considerable tweaking may be required to get a similar result in SilverFast.

Quote
For me there are three tell-tale signs of good colour in a prescan: the grey balance for tones that I know are supposed to be grey, the hue of the sky and the colours of objects that I remember very well. Between one or more of these three clues, I have found it quite manageable to lock into believable colours at the scanning stage, usually with just a couple of grey "pipette" placers on the image.

Quite so. The problems arise when there are no identifiably neutral grey areas in the image, no blue sky and no green foliage etc and the negative might not even be your own and/or your memory does not serve you well. (I've got those skin tones looking just right but wasn't that bikini top more blue than cyan, weren't the walls cream rather than pale green etc?

With Vuescan and negatives I can usually rely upon a credible balance from the start even though I might not be able to get, say, the full saturation I want due to the minimal controls.

Quote
Because ARGB is a smaller colour space than ProPhoto it stands to reason that images could be overall less saturated in the former than the latter, but not necessarily - depends on the image.

That's no quite what I mean and this issue highlights yet another oddity which causes me to wonder if there's something broken in the color management system with regard to Negafix. For example, I prescan a positive with CMS set to embed the ProPhoto profile. After editing so the preview looks just right, I change my mind and decide I want to embed the ARGB profile. (Or maybe I'd just remembered CMS was set to the wrong profile.) I can go back into 'Options', change ProPhoto to ARGB. I get a notice advising me to do another prescan, but either way the preview remains unchanged in appearance. Sometimes, for a brief fraction of a second I see the colors in the preview suddenly become more saturated as I make the change, probably because I've got other programs running in the background which are slowing my computer.

It's clear that SilverFast does a profile conversion in this situation and changes the numbers. Both the ProPhoto and ARGB scans from the same preview (or at least the same adjustments to the preview which has had a fresh prescan) look identical to my eyes. To test this with regard to skin-tone reds, I scanned a close-up mug shot of myself wearing a deep pink shirt, taken on Fuji Provia slide film in bright daylight. What I did notice is that in both cases the deeper reds in the shadows were very, very slightly more saturated in the preview than in both scanned images in Photoshop. In other words, there was a very slight mismatch between the preview and the scanned image which was barely discernible (and so slight as to be of little concern to me), but no discernible difference between the ARGB and ProPhoto scans.

However, what I do find a bit of a worry, is that the above situation does not apply when scanning negatives, and I'm not sure if it's supposed to be like that. In other words, SilverFast does not do any profile conversions to the edited preview if one decides to change the profile later in 'Options/CMS'. To compare ARGB and Prophoto scans from the same preview, one has to assign the ProPhoto profile to the ARGB image in Photoshop, or vice versa, and then both images look identical.
Logged

Jonathan Wienke

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5829
    • http://visual-vacations.com/
What's wrong with SilverFast?
« Reply #3 on: September 25, 2005, 10:48:32 pm »

Quote
However, what I do find a bit of a worry, is that the above situation does not apply when scanning negatives, and I'm not sure if it's supposed to be like that. In other words, SilverFast does not do any profile conversions to the edited preview if one decides to change the profile later in 'Options/CMS'. To compare ARGB and Prophoto scans from the same preview, one has to assign the ProPhoto profile to the ARGB image in Photoshop, or vice versa, and then both images look identical.
My guess is that that is the root of the problem, and Mark is scanning negative film. After looking at the scan crops he sent me, I think a good case can be made that no matter what color output profile is selected, the same RGB data is output and is then tagged with the selected output profile. Please note that all of the images below have been converted to sRGB so that they display quasi-properly on the web. Mark sent me crops of two images straight from Silverfast, the first with Adobe RGB selected, and the second with ProPhoto. There was a minor levels difference between them, so I converted them both to 16-bit, did a radius 1 Gaussian blur to smooth out the CCD noise/film grain, and did identical auto-levels on both images to equalize the white and black points, with the following results:

 

Note that the ProPhoto version clearly exhibits the over-saturated reds Mark was complaining about.

Now for comparison, here are the same two images with the profile assignments swapped:

 

I'd say this is a fairly convincing demonstration that with only minor differences, the same RGB data is being output no matter what output profile is selected, and then the color profile is assigned as an afterthought. Certainly the proper "assign the film profile to the RAW scan data and then convert to the selected output profile" procedure is NOT being followed, at least with Mark's current settings. Not being a Silverfast expert, I'll refrain from theorizing whether this is due to a configuration error or a program bug, but certainly good color management practice is not being followed by Silverfast here.
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
What's wrong with SilverFast?
« Reply #4 on: September 26, 2005, 01:01:04 am »

Jonathan,
Thanks for posting those illustrations. Those differences are exacly the sort of differences I get when comparing negative scans of the same edited preview with the 2 different embedded profiles. What we see here are the different renderings by ARGB and ProPhoto of the same numbers. However, these differences do show up in the SilverFast preview (with negatives) if one switches profiles before scanning. It doesn't have any bearing on the fact that whatever profile is chosen to be embedded, the scanned result when opened in PS, preserving the embedded profile, will (should) look the same as the preview.

Before I contact LaserSoft about the apparent uselessness of their negative film profiles (with recently shot film, not just faded stuff), I'll try loading the Win2K version of SilverFast on another computer to see if it performs any differently.
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
What's wrong with SilverFast?
« Reply #5 on: September 26, 2005, 10:50:35 am »

Ray, I believe the differences between the two models of the scanner are not only confined to the light source - which is probably not a determinative issue - but more importantly to software and firmware as well. Minolta have claimed that model 2 is more adept than the first model at handling colour negative media, and I believe this was confirmed in one or another review I do not remember to specifically identify. So I am not surprised that an owner of model 2 would be having more success at negatives with Dimage Scan than did people like me with the first model. (Not that it was "unsuccessful" - just that there was alot of post-scanning adjustment needed to make the image look right.) The reason I ranked Silverfast first and DimageScan last with Vuescan in-between is that for ScanElite 5400-1 and the negative material I'm using (a brand of Fuji Superia ASA 100), that is simply how the results came out, the assessment criteria being two things: (1) flexibility and scope of allowable adjustments in the scanning software itself, and (2) the resulting correctness of luminosity, hue and saturation of the scanned images, as measured by how much or how little post-scanning adjustment is then required in Photoshop. The only major exceptions to my overall preference for Silverfast are (1) their dreadful interface and documentation and (2) this reddening business I hit upon and raised in this Forum.

In the version of Silverfast that Lasersoft markets for the first model of the scanner, it is not necessary to repeat the pre-scan when you change the colour space in the CMS (colour management system) dialogue window. But in order to see the effect of the change of colour space in the pre-scan window it is essential to click the "APPLY" button - otherwise the change does not take hold and you will see no difference.

The exercise in which Jonathan and I have engaged over the past couple of days does raise one or two fascinating - and perhaps troubling - issues about which I want to think  more before commenting on the conclusions that Jonathan has reached. I confirm that the images he posted closely resemble the output that I sent him notwithstanding the conversion to sRGB and the few tweaks he mentions - the problem is correctly displayed.

I would also like to confirm now - because this needs no further thinking - there is nothing amiss in the manner that I have specified the set-up for using Silverfast with negatives. Ray, you and I have been the whole-nine-miles through many of those set-up options, I have studied the Silverfast Manual, Ian Lyons' tutorials and Taz Tally's book and experimented sufficiently with all the settings in CMS and Negafix that I have confidence in the appropriateness of my settings. I have no confidence, however, - because I don't know - what Silverfast is doing with those settings! And I continue to believe there is utter confusion in how Lasersoft describes and represents the whole colour management aspect of their software - and I have told them as much. (That however may be only a presentational issue and not necessarily programming defects, but I'm not sure we know for sure yet.)

Hence, if there is a problem of the kind that you and Jonathan are pointing to, it is either a generic issue about the choice of colour space in some conditions, or the kind of defect in Silverfast you two are alluding to. That is the aspect I want to comb through some more, and I shall report back any further views I have on it. I was also considering approaching Lasersoft tech support once again - despite the possibility that if they see another email from me they will probably want to tear-out either their own hair or mine!

Meanwhile, I wish to thank you and Jonathan ever so much for the high level of professional collaboration on this interesting question. It really makes it worthwhile to participate in a Forum that facilitates this level of dialogue and problem-solving.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Jonathan Wienke

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5829
    • http://visual-vacations.com/
What's wrong with SilverFast?
« Reply #6 on: September 26, 2005, 11:54:44 am »

Quote
I confirm that the images he posted closely resemble the output that I sent him notwithstanding the conversion to sRGB and the few tweaks he mentions - the problem is correctly displayed.

I would also like to confirm now - because this needs no further thinking - there is nothing amiss in the manner that I have specified the set-up for using Silverfast with negatives.
If that is indeed the case (and if you're on a first-name basis with LaserSoft's tech support I see no reason to doubt this) then Silverfast indeed has a problem with color management when scanning negatives.

Quote
Meanwhile, I wish to thank you and Jonathan ever so much for the high level of professional collaboration on this interesting question.

You're welcome.
Logged

imdave2

  • Newbie
  • *
  • Offline Offline
  • Posts: 3
    • http://imdave2.shadesoflilly.com
What's wrong with SilverFast?
« Reply #7 on: September 26, 2005, 03:28:01 pm »

Question for Jonathan:

Back in the 8/16bit thread, from where I think this thread originated, you mentioned the following:

"Good scanner software will have a color profile for the film stock you're scanning, and will assign the film profile to the raw scan data"

Could you provide an example of "Good scanner software"? I have only used the included software with consumer level scanners (Nikon, Epson), as well as Silverfast. Like everyone else, I find color negatives to be a challenge.

Thanks, Dave
Logged
~ Dave ~
imdave2.shadesoflilly.com

Jonathan Wienke

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5829
    • http://visual-vacations.com/
What's wrong with SilverFast?
« Reply #8 on: September 26, 2005, 07:12:21 pm »

I haven't scanned film in over two years, and have no plans to do so any time soon. Your question would be better answered by someone who actually scans negatives regularly.
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
What's wrong with SilverFast?
« Reply #9 on: September 26, 2005, 07:20:36 pm »

Imdave2, I do scan negatives regularly, and yes you are correct about there being a challenge. I have a very practical way of addessing your question: assuming you own a scanner that can use these softwares, download each one of them (the software native to your scanner brand, VueScan and Silverfast - which you say you have already tried) as demos and work with them for a variety of images, then adopt the one you think gives you the best results with the least trouble. Because it is so hard to know what is "under the hood" with any of these products, the empirical way is the best - and after all, it is only the results that count.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
What's wrong with SilverFast?
« Reply #10 on: September 26, 2005, 10:10:28 pm »

Mark,
The basic issue scanning negatives with SilverFast, for me, is the fact that the initial appearance of the preview seems way off, despite having chosen the supposedly correct film type. One has to work on it, but having worked on it, the result can then be more satisfactory, more accurate if your memory for color schemes is good, than the less complicated scan using Vuescan.

There's no doubt that the color adjustment tools in SilverFast are very powerful and therefore, ultimately, with perseverence and a bit of practice, I have no doubt that better results can be achieved, whether one is scanning slides or negatives.

Installing SilverFast on another machine and operating system and starting off from scratch as it were, I got the impression the film profiles in Negafix have been built with Adobe RGB in mind. Use a wider gamut space such as ProPhoto or Wide Gamut RGB and the starting position, the prescan, is off by a massive degree. Is this what you find?
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
What's wrong with SilverFast?
« Reply #11 on: September 26, 2005, 10:53:25 pm »

Quote
I got the impression the film profiles in Negafix have been built with Adobe RGB in mind. Use a wider gamut space such as ProPhoto or Wide Gamut RGB and the starting position, the prescan, is off by a massive degree. Is this what you find?
Ray, this is partly the issue you and I, and I and Lasersoft Tech Support spent so much time resolving before the Discussion Forum crash. The conclusion we came to, in part as a result of one of your suggestions, was that the following combination of CMS settings provided a close match between the Preview Scan and how the file looks once opened in Photoshop: Internal to Monitor: use ICM; Internal to Output: use RGB; Under Profiles for ICM: Internal: use ProPhoto RGB, Grey>None; Rendering Intent: Rel Colorimetric; Embedded ICC profile checked. Profile to embed: Silverfast automatically inserts the choice made at Profiles for ICM (in this e.g. ProPhoto).

The above of course only deals with CONSISTENCY between what you see in the Prescan and what opens in Photoshop, and as far as I'm concerned, that aspect of the process we have jointly resolved.

The other aspect your question raises is whether the colours and luminosity we see in the Prescan are "correct" (by which I mean, believable, what we remember of the scene, neutral greys where they should be, no purple cows, etc.) Getting this aspect right I find depends on essentially three things that are easy to manage: (1) for negatives, the most appropriate choice of the negative film in Negafix, (2) again in Negafix, tweaks of exposure and "auto tolerance" (a dumb name for an ill-defined concept) and (3) use of the grey "pipette" obscurely represented as an eyedropper (nothing to do with grey! ha!) in the middle of the "Highlight/Shadow" tool that sits on top of the "General" and "Frame" tabs. After just a bit of experience, using these three key instruments is generally sufficient to produce a pre-scan that is uncannily correct for images that aren't very under or over exposed.

The "L" or "N" adjustment in what they label as "Gradation adjustments" and/or the Exposure control or Midtone Brightness control in "Picture Settings" are about the best they offer for rescuing severely over or under exposed images. The rest of their curves-type adjustments are far less useful than Photoshop's, so I don't bother with them. Very infrequently I may resort to the Global Color Correction pie, which can be useful for severe cases of undesirable colour cast on images where there is nothing neutral grey on which to focus the grey pipette.

So the bottom line: the above recipe gives me images in Pro-Photo colour space that are (a) quite accurate, and ( consistent between PreScan and Photoshop.

But to every generalization there is an exception, and here we go: The troublesome issue remains that maddening red skin business that Jonathan and I were working on a bit earlier in this thread. He thinks it is due to crummy colour management processes within Silverfast, and I think maybe he's right but I'm not 100% there yet. In fact, this issue is related to another issue I raised in another thread about whether Epson 4000 printers can see red; here I also discovered that narrowing the color working space from ProPhoto to ARGB98 in Silverfast ameliorates this problem too (keeps the reds more - but not entirely - like what they should be when you soft-proof in Photoshop). So the general exception to the high quality of Silverfast working in ProPhoto is that it tends to blow certain reds out of gamut producing - depending on the image and the colours involved - either impossible skin tones or shades of red that Photoshop can't squeeze into the printer's gamut in a nearly believable way. The one thing I haven't tested for this situation - and in principle could have a positive impact - is to change the rendering intent in Silverfast CMS to "Perceptual". I shall likely try that tomorrow and post my findings.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
What's wrong with SilverFast?
« Reply #12 on: September 27, 2005, 01:03:28 am »

Quote
Quote
The troublesome issue remains that maddening red skin business that Jonathan and I were working on a bit earlier in this thread. He thinks it is due to crummy colour management processes within Silverfast, and I think maybe he's right but I'm not 100% there yet.

Mark,
You seem to be very familiar with the SilverFast tools so I'm still puzzled as to the nature of the problem. Provided the scanned output matches the edited preview, and considering how powerful those color adjustment tools in SilverFast really are, the only problem I can see are:-

(1) The amount of work involved in getting the preview into shape.

(2) It's much easier for colors to be credible but not accurate than it is for colors to be accurate but not credible.

I also don't use the 'global color correction tool' much. The 'grey balancing' pipette is invaluable, but also the 'Selective Color Correction' tool, (that circle with 6 complementary colors on the circumference and the primary color in the centre that mostly closely represents the color in the image one has clicked on.)

The excessive redness in skin tones that is always apparent in the preview with negatives if one is using ProPhoto, can be easily tamed with 'slective color correction'.

This is the procedure. Open SCC and click on an area of excessively red skin in the preview. The small ball in the centre of the large circle turns red, the closest primary to the red skin. Move either (or both) the small magenta and/or red balls on the circumference towards the central red ball, one at a time, by degrees. Movement from an outer ball towards the centre reduces that color component. Movement from the inner ball to any outer ball increases the color component represented by the outer ball. The direction of the movement is indicated by an arrow. As soon as you stop the movement, the change of hue becomes apparent.

You can move from the inner ball to any of the 6 outer balls or vice versa. You can become a juggler  . This tool, in my view, is far more powerful and useful than Photoshop's Hue/Sat control. It basically allows you to get any color, shade or hue you might desire in any part of the image. Unfortunately, there is some slight 'spill-over'. Each adjustment is not 100% localised so, for example, with a bit of work it's not too difficult to get a 'perfect' skin tone, but the background wall might then be pale cyan instead of the true pale cream that one remembers. Of course, if one can't remember what color the background wall really was, one can always check by scanning the negative with Vuescan  :D .
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
What's wrong with SilverFast?
« Reply #13 on: September 27, 2005, 08:35:44 am »

Ray, re your two problems:

(1) The learning curve is a bit steep because the documentation and the interface are crummy. But once down that curve, routine use of the program to the get preview scan into good shape is actually quite easy.

(2) This is a deep subject. After long and hard thought and observation about this, apart from laboratory conditions that are so constructed (such as photographing color checkers or product photography, where the glass of beer must be made to look EXACTLY like the glass of beer would look in "reality" - and even here under what lighting conditions for the real thing and the ad?) there really is no such thing as "accurate" colour in most photography. It is in the mind's eye. Furthermore, most people would be incapable of knowing it when they see it unless there were comparative material available (e.g. when I do some product photography for my wife's design business I see the product and the photo side by side, so we can come close). That is why we depend very much on neutral grey, black point and white point to be our colour balancers for that vast majority of photographs we made some time ago and the precise colours of this and that are a fog. For example, I am scanning negatives of photographs I made at the Temple of the Reclining Buddha in Bangkok about two years ago. Do you think I really remember the exact hue of the gold leaf paint on the Buddha? No way. One really does depend on colour balancing guidelines and what looks believable and reasonable to the mind's eye.

Thanks for the suggestion on using SCC. I know the tool in Silverfast, as well as those in Photoshop, but I look upon that as a last resort - the reason being that often if you think you need to use this, it is a sympton of a more generic problem - so first try to resolve that - and if still needed thereafter, use SCC. SCC is also useful, of course, for intentionally changing a colour from what it "should be in reality". And re-scanning the negative in Vuescan  :D  is no assurance of anything - just more information to hem-and-haw over!
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
What's wrong with SilverFast?
« Reply #14 on: September 27, 2005, 10:33:35 am »

Mark,
Your above points are largely true. At least I wouldn't argue with them in general terms. The gold leaf doesn't need to be the exact shade that you saw in the temple in Bangkok as long as it's recognizable as gold leaf, and the same applies to the shade of many other things that we are familiar with, in particular the complexions of people.

I can now see that a large part of my initial disatisfaction with the way SilverFast handles negatives is due to the fact that the film profiles in Negafix are simply not accurate in the ProPhoto RGB color space. Because most of my scanning has been of slides where it makes no difference to the over all color balance whether one chooses Argb or ProPhoto, I had assumed the same would apply scanning negatives. It clearly doesn't and that's the source of the confusion.

Since there seems to be little point in choosing a color space that requires one to make more substantial color corrections than necessary, it would seem to be more sensible to scan all negatives into Adobe RGB when using SilverFast. Problem solved  :D .
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
What's wrong with SilverFast?
« Reply #15 on: September 27, 2005, 08:27:30 pm »

Reverting to Jonathan's comparison of the four images several posts ago in this thread. Jonathan may well be correct that the test shows the scanned data is the same whatever the selected colour working space, and the file is simply tagged with the work space one selects; but it isn't clear to me how that explains why it gives such incorrect colour information for reds with the ProPhoto tag. I shall raise it with Lasersoft, because I think it does require knowing more about the guts of the program.

Ray, I have successfully scanned negatives that do not have brilliant reds or skin tones in ProPhoto colour working space; but you are correct that scanning in ARGB98 is pretty-well failsafe regardless of the image. However, doing that sacrifices some hues that an Epson 4000 can reproduce. Perhaps this is the lesser of two evils. I would like to do some more testing of the ProPhoto issue and will revert with any results – probably several days – other stuff on the front-burner just now.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
What's wrong with SilverFast?
« Reply #16 on: September 27, 2005, 09:39:59 pm »

Ray, Jonathan, here is the text that I sent to Lasersoft technical support this evening. I shall let you know what they reply to me. By the way, if you think this needs to be amended or amplified please let me know - I can get back to them.

<<It seems there are still problems with ProPhoto work space used in Silverfast - this issue somewhat different from the one we were discussing previously for which I informed you about the solution. When scanning certain values of red in textiles and fleshtones using Silverfast in ProPhoto working space, the image file turns out to be a highly saturated version of red skin tones that is completely un-natural, and for the other reds - out of gamut - unprintable. When the same image is rescanned in Adobe RGB98 colour working space, the reds come back much more under control - there is far less saturation of skin tones, and the other reds are more correctly handled in the printer. Diagnosis done by members of the group working on this problem suggests that Silverfast does not actually convert the data to the working space as part of the scanning process, but only tags the scanned data with the colour working space information. That is to say, no matter what colour working space one chooses in Silverfast, the scanned data is the same for all - only differing by the tag for the chosen work space. Is this correct? Those of us who have been working with this problem on an Internet Discussion Forum need your advice on exactly how Silverfast converts data as a function of the chosen working space, and whether indeed Negafix is intended to work properly with ProPhoto colour space. The standing impression so far is that Silverfast cannot properly render colour in Pro Photo colour space. I shall be grateful for whatever advice you can provide on this issue. For your guidance I am attaching two images that demonstrate the problem. The image file names tell you which is which. Please make sure when you open them you conserve the embedded colour working space so you will see the difference.>>
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
What's wrong with SilverFast?
« Reply #17 on: September 28, 2005, 01:37:26 pm »

Ray, Jonathan, in what was a stellar turn-around of emails between Lasersoft and me, I have just now received their reply to the enquiry which I sent them last night and posted immediately above/below (however the thread is ordered). I believe this issue is now adequately de-mystified. Here is the response from Lasersoft:

<<ProPhoto works fine if there is also a scanner profile. If you scan a slide, have your IT8 profile activated, the scanner profile will be used as a reference, then the image will be converted into ProPhoto. The image looks ok, not oversaturated.

If there is no input profile available, because you are scanning a negative, or because you don't have IT8 calibration, there is no actual conversion into ProPhoto, because no reference is available for the input data. The output is being tagged as ProPhoto and this results in the highly saturated images.

Adobe RGB1998 closer resembles the actual range of the scanner. Therefore these problems do not occur (or more specific: are not as noticeable) versus using ProPhoto. Apparently you don't gain anything in terms of quality by using ProPhoto. The image data conformably fits into Adobe RGB- Prophoto uses just a different matrix.

If you scan the image with no profiles used whatsoever, and then convert to ProPhoto using Photoshop, you should be getting the same saturation problem.>>

You will notice his reply confirms for negatives a behaviour which you and Jonathan have suggested, and also his very last sentence confirms a professional view (which one member of L-L debunked as "malarky") that the colour space itself could be a contributing factor to excessive saturation.

This last point is interesting. I wonder if it happens because of the re-mapping of RGB values further up the saturation level going from a narrower (viz ARGB98)to a wider (viz ProPhoto) colour space. Does that sound right?
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Jonathan Wienke

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5829
    • http://visual-vacations.com/
What's wrong with SilverFast?
« Reply #18 on: September 28, 2005, 02:54:08 pm »

Quote
You will notice his reply confirms for negatives a behaviour which you and Jonathan have suggested, and also his very last sentence confirms a professional view (which one member of L-L debunked as "malarky") that the colour space itself could be a contributing factor to excessive saturation.

This last point is interesting. I wonder if it happens because of the re-mapping of RGB values further up the saturation level going from a narrower (viz ARGB98)to a wider (viz ProPhoto) colour space. Does that sound right?
Not quite. The RGB values output stay the same, but the way they're interpreted changes depending on the profile attached to them. The short version of what you were told is that when scanning negatives without a custom profile, no color space conversion is attempted and the scan data is simply tagged with whatever output profile you select. So the solution is to make/buy/steal a custom film profile to allow the color space conversion. But why the generic film profiles don't count when scanning negatives is beyond me.
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
What's wrong with SilverFast?
« Reply #19 on: September 28, 2005, 04:03:55 pm »

Jonathan, thanks for the clarification on the process that's happening. I've consulted with a number of specialists in this area - Wolf Faust, Don Hutcheson (Hutchcolor), CF Systems, looked-up Blatner/Fraser, etc. etc. and the deal is that no-one makes ICC profiles for colour negative materials because of the huge variation in film types, each with their version of orange mask, and other factors Hutcheson describes in a paper, essentially saying that negative materials defy profiling. So they are simply not available. That is why Silverfast has this Negafix module which allows you to select your film type, tweak it, and then it adjusts the scanned data accordingly in the process of converting it from negative to positive. So it is by nature an ex post facto process that kind of simulates the behaviour of a profile without being one.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."
Pages: [1] 2   Go Up