Luminous Landscape Forum
Raw & Post Processing, Printing => Digital Image Processing => Topic started by: chaddro on November 02, 2012, 06:36:24 pm
-
Hi All!
I am scanning a number of stamps from the 20's and 30's for a client project. The paper is quite matte and there is no shine to it. As a backer for the stamps (they've been removed from the envelopes) I'm using a white matte acid free cotton paper so they look mounted (well, sort-of!).
One of the problems I'm having is at the leading edge (top of the stamp) I'm getting blown highlights. And histogram is piled up against the right.
I have tried quite a few options from within Silverfast AI 8 to fix this, turned off most/all of the auto adjustments, but still can't seem to pull the exposure down.
Yes, there is an "Exposure" slider, but all it does is pull the data to the left, it doesn't actually "stop down" the scanner so I don't get an over exposed scan.
I'm hoping someone can point me in the right direction. Really banging my head against the wall on this one.
Thanks!
-chadd
Mac OSX 10.6.8 + Epson V700 + Silverfast Studio Ai 8.0.1.r17
-
Hi Chadd,
I just tried replicating your problem, as I have the V750 scanner (same profiles and scanning specs) with SF8 Ai Studio. The default scan produced an overexposed prescan, highlights fringing on clipping, with little to distinguish the three different shades of white (see illustrations) and histogram values bunched to the right, I suppose like you got. I opened the histogram tool, Expert dialog, clicked on Color Cast Removal and made sure the slider was at 100. This solved the problem. Clipping disappeared from the output histogram, (but the input histogram did not update). No matter, the Densitometer shows that there are no clipped values and the three shades of white are clearly distinguishable in the adjusted settings. Give it a try. You can also go into "Preferences> Auto" and shift the "Auto-adjust Darker" slider rightward and see to what extent that helps.
Scanning applications don't have highlight recovery like we have in Lightroom or Camera Raw, where rather than just turning clipped values from white to gray as you reduce exposure, the Adobe applications simulate RGB values having colours. If you can succeed in getting the scan values below level 255 in all three channels or even in two of three, make the scan, open it in Lightroom and try additional highlight taming there.
-
Hi Mark!
Thank you for your valuable input. The Color Cast Removal at 100 checked did help. I also gave the Auto Adjust Darker slider a value of 5. I'm now getting a very good pre-scan.
The Auto Adjust Darker/Lighter sliders were both set to 15. Perhaps that was also complicating things.
A couple more questions for you if that's okay...
1) Is it better to adjust the image after scan, or okay to tweak the Output Histogram in Silverfast? My goal would be to bring out the texture of the paper a bit more.
2) From your experience do you think that Lightroom 4's sharpening is superior to Photoshop CS 5's sharpening? I have both.
I've been asked to make 30" canvas prints of these 1" high stamps. The stamps are illustrative line art so I feel it's quite doable.
I'd be also interested on your thoughts of sharpening something like this.
For myself, I've been converting to LAB and sharpening on the L channel with very acceptable results.
Thank you again for your suggestions!
-chadd
-
Hi again Chadd,
Glad that all worked.
The workflow questions you ask are perennial, there isn't one unambiguous answer and as usual - "it depends". I go into a lot of this in my book on SilverFast, so if you wish to explore this topic in more depth I'll be unbashful enough to recommend my book to you. Click on the link under my signature. But let me try to respond here and now to the extent reasonable: My general approach is to get the scan as close to a finished product as the scanning software allows. That way, you bring it into a post-scan image editor in pretty decent shape, and that gives you more degrees of freedom for further adjustments while preserving the basic quality and integrity of the image. So if you find that making luminosity adjustments in SilverFast improves the appearance of texture of the paper in the SilverFast Prescan, by all means do it. It's just important to bear in mind that a scan bakes in whatever you scanned. Also, because you intend to make very big enlargements of these scans, use the scanner's highest optical resolution and largest scan dimensions you can before the Res slider in the image dimensions panel departs from the green zone. Use this panel in "Expert" dialog mode. Also make sure to scan in 48-bit RGB mode.
As for sharpening, I'm a big fan of two sharpening tools that I think are really state of the art - Lightroom capture sharpening in the Detail panel, but if working in Photoshop, the Pixelgenius Plug-in Photokit Sharpener 2. If you intend to sharpen in either of these applications, do not sharpen in SilverFast. Also, either of these tools completely obviates the nuisance of Lab conversions and they work superbly well. While on this subject and going back to your objective of making big enlargements that well show the paper texture, you may find that Lightroom's Clarity tool, as well as its Contrast tool would make a useful contribution. You can play non-destructively with various combinations of Clarity, Contrast and Sharpening to get the effect you like. Also within Photoshop, the Photokit Sharpener 2 package has Creative Sharpeners that will help a lot. The key, however, is to make sure the pixels get scanned-in at high resolution with no clipping.
-
Chadd, for comparison, here's a copy of Mark's first, unadjusted stamp that I edited very quickly in LR to bring out the paper texture.
-
Just a thought for the OP: what profile are you using for the scanner? If the white point of the target used to generate the scanner profile is not as bright as the white point of the stamp then the profile, once assigned, will clip the whites of the stamp.
Solution? use a target with a very bright white point (like a Colorchecker SG) to generate the profile. Targets printed on regular photographic paper (i.e. fuji crystal archive) don't have very bright white points.
I also have a V700 and had this problem for a while with scans of watercolour paintings. Some watercolour paper is seriously bright and I was losing the paper colour and the faint colour washes. My first target was an HCT, which has a low white point.
-
Chadd, for comparison, here's a copy of Mark's first, unadjusted stamp that I edited very quickly in LR to bring out the paper texture.
Good job Dean, and shows what lies "under the hood" from that scanner and software once it is brought out properly.
-
Just a thought for the OP: what profile are you using for the scanner? If the white point of the target used to generate the scanner profile is not as bright as the white point of the stamp then the profile, once assigned, will clip the whites of the stamp.
Solution? use a target with a very bright white point (like a Colorchecker SG) to generate the profile. Targets printed on regular photographic paper (i.e. fuji crystal archive) don't have very bright white points.
I also have a V700 and had this problem for a while with scans of watercolour paintings. Some watercolour paper is seriously bright and I was losing the paper colour and the faint colour washes. My first target was an HCT, which has a low white point.
The first thing I verified before posting a response to the OP was the colour management settings in Preferences>CMS. I determined that the software was pulling the correct profile for reflective scans using this scanner class. That of course is a canned profile provided by LaserSoft Imaging (LSI) based on their profile-making using their unit of that scanner model. The quality of the canned profile for the OPs' particular scanner then depends on any performance differences between his scanner and the scanner LSI used for generating the profile. As Epson manufactures these professional units to a high standard of uniformity (judged by lack of evidence to the contrary), the normal expectation is that the canned profile should do a pretty good job on most of the users' units.
However, in case it doesn't, as you may know, SilverFast 8 Ai Studio does have an Auto-IT8 profiling capability, and LSI makes available their reflective IT8 target for users to generate their own profiles. My copy of this target was printed on Kodak Professional Endura. To test the difference between the canned profile and a custom profile for my scanner, I created a custom profile with it. It turned out that in terms of white point and white balance, the custom profile and the canned profile are very similar - not the same but close. The main difference was a correction of 2 levels in the a* and b* channels shifting the colour balance from very slightly bluish to a more neutral rendering. In both cases the measured white point picked up from the LSI IT8 target is L90. "Maximum white" would be L100, and middle gray L50. The fact that a profiling target itself does not contain a patch having L* closer to 100 does not necessarily mean that any brightness values exceeding the maximum L* of the profiling patch would be clipped. That very limited universe of patches on a profiling target merely provides the sampled base data or anchors for algorithms in the computer's CMM that use this information to calculate millions of tones and colours that will be all over the place - above, below, and beside the values of individual patches in a profiling target. The better success you obtained making your own profile is most likely due to the possibility that your custom profile better replicated the behaviour of your scanner than the profile you were using previously.
All that said, not impossible that in principle profiling and CM performance can have something to do with clipping image data, but in this case, especially in light of what Dean and I did above, I doubt very much this is the issue.
-
The better success you obtained making your own profile is most likely due to the possibility that your custom profile better replicated the behaviour of your scanner than the profile you were using previously.
All that said, not impossible that in principle profiling and CM performance can have something to do with clipping image data, but in this case, especially in light of what Dean and I did above, I doubt very much this is the issue.
Your last paragraph made me read the first two posts again, a little more carefully this time. I guess I just reacted to the concept of his problem and threw out a suggestion based on my experience.
I must re-iterate, however, that I was comparing two custom profiles: one made with the HCT, one with the DCSG. HCT profile gave me clipped highlights, DCSG profile contained them.
I could extend the HCT profile by fiddling the white point of the target scan, but this gave me extra work to do with exposure and contrast correction. The DCSG target with its brighter white point was just a neater solution.
-
Thank you both for you additional input. I did try messing around with the CMS settings to see what they would get me, but the default canned reflective profile seemed to do better than anything else I tried. Turning off all the CMS setting really bunched the histogram up to the right side.
I do have a Digital ColorChecker SG, but I hadn't thought to try to roll a custom profile for the v700. I had looked at the it-8 features of Silverfast long ago, but really don't know which target to get or if I really needed one... This was the first time I had this kind of trouble with me scanner.
Dean, does your e-book cover scanner profiling?
I ended up using Photoshop for sharpening using LAB and the L channel method. I'm sure Lightroom can produce superior results, but my first attempts weren't going well and I absolutely had to finish printing last night (finished at 3 am!). I can spray coat today and deliver tomorrow, so I'm really grateful for the timely help!
Stephen, I'd be interested in your workflow for creating profiles for the v700 with your DCSG.
Again, thank you both for your help.
-chadd
-
I don't know that Dean has an e-book, but my e-book does. However, the profiling discussion focuses on using SilverFast 8 and its targets.
-
I must re-iterate, however, that I was comparing two custom profiles: one made with the HCT, one with the DCSG. HCT profile gave me clipped highlights, DCSG profile contained them.
OK, I can see where you are coming from - that DCSG is a pretty high-end piece of equipment and likely the profiling software you are using as well. It's therefore reasonable to expect you are getting higher quality profiles. Interestingly, the steps I recommended to Chadd appear to have retrieved enough highlight information to be usable, so again, I'm just not sure that profiling is his issue, though for sure worthwhile exploring.
-
My V700 workflow is pretty simple:
I use Vuescan and output it's 'Raw' scans. Basically a linear tiff, straight off the scanner, no adjustments of any sort. Don't know if Silverfast can do this, maybe its profile generation uses a similar raw scan. Mark? Also: can Silverfast make use of the DCSG as a target?
I used Profilemaker to generate the profile from a raw scan of the DCSG. (Didn't buy one myself, just rented it for a day about $20 equivalent and have been using the profile very happily for ages now.) Free alternative to PM would be Argyll, but needs lots of messing around before you get good results.
Assign profile to scans made in exactly the same way as the target scan. Tone and, depending on the pigments being scanned, colour pops nicely. 95% of the work done. One or two hard proofs and minor tweaks get an excellent match.
Something I just stumbled across two days ago: CoCa (http://www.muscallidus.com/coca/) It's a very friendly front-end for Argyll camera profiling but you can use it for scanners too, if you select the right algorithm. I played with it for a few hours on Friday morning and early results suggest that the profile quality is better than PM5. Haven't used a CoCa profile on a watercolour painting job yet. For the DCSG I had to use an Argyll-friendly version of the reference file that comes with PM5 here (http://stephensarchive.webs.com/misc/ColorCheckerSG.cie) I used the tiff from here (http://www.libraw.org/node/64) to compare profiling results. just layered them and used difference blending.
-
Hi Stephen,
Yes, SilverFast has been able to output a "raw" scan for a very long time now, in the sense of a linear TIFF with no adjustments made. You can use any scanner profile you wish as the Input profile with SilverFast; however SilverFast' Auto IT8 profiling process will itself only self-generate profiles using LSI's supplied targets, which are bar-coded and accompanied with their own reference files. Using this process one doesn't need other targets or profiling software, but of course it is limited to their own materials.
The procedure you describe is obviously a lot more work, but very interesting if it's delivering better profile quality that makes a difference for certain kinds of output.
-
Stephen, recently I also tested CoCa to profile my scanner using its 16 bit linear output. See this thread: http://www.luminous-landscape.com/forum/index.php?topic=71953.0
Which algorithm did you find worked best for your scanner?
-
It was one of the 'clut' algorithms, I think, but can't recall right now. The tests and profiles are on my work machine which I'll get back to tomorrow morning. (about 15 hours from now. I'm on GMT+2)
Like I said I've only played with it for a few hours - very limited testing, but it does look very promising.
-
Stephen, please keep us informed on your progress with CoCa.
I just created another profile with CoCa. This time I used the LAB algorithm, whereas before I used XYZ. I started with the same scan of my Provia IT8.7 target, used the same target reference file, and all the other same settings in CoCa, except to change the algorithm. Once again, however, as described in my the thread I linked to above, the profile was exactly the same as all the others I created with CoCa. Something is seriously amiss, but I have no idea what. CoCa is so easy to use, but maybe something I'm doing is messing up the profiles, or maybe something is screwed up with CoCa. Very puzzling.
EDIT: I just made a similar profile using the Gamma-Matrix algorithm. This one is slightly different than the others. So, maybe I'm actually doing everything correctly?
-
This is very interesting... I've played with Argyll some, but since I've been mostly using Epson's paper's with their canned profiles I've been pretty happy. I'll take a look at Coca too.
Here is one of the stamps I was working on. Original size, and a crop from the large print. I think my client will be pretty happy with the results. It was just crazy wanting to make a 30" prints from 1" high stamp but I think I pulled it off. Really appreciate the help here. My post over at Siverfast still has been approved. No wonder their forums are so dead.
Mark, is there a general purpose IT-8 target that you would recommend? Do I need different targets for reflective and negative profiling?
I have a big personal project coming up. About 300 negatives from some 50+ years ago. The prints were destroyed in a fire. It is a large mix of 35mm color, b/w and these large square negative (mostly b/w).
Stephen, I do have PM5 so I can give the DCSG a go and try my own profiles. One question though: does the profile replace the scanner profile in silverfast, or does it get applied to the image in photoshop after scanning?
-chadd
-
Hi Chadd,
Firstly - remarkably good results. I'm really pleased you got that detail and clarity as well as you did. It speaks well for both the scanner and the operator.
I guess you meant to say your post at SilverFast Forum has NOT been approved?
There is no such thing as a general purpose IT8 target. There are Kodachrome targets, non-Kodachrome positive transparency targets and reflective targets, each for their own media type. Kodachrome targets by the way are just about impossible to find outside LSI and they are getting scarce there too, because the stock cannot be replenished. No more Kodachrome and no more Kodachrome processing labs, but no supply problems for non-Kodachrome targets, be it from LSI, Don Hutcheson or others.
There is no such thing as IT8 scanner profile targets for negatives. For negatives, in SilverFast you would use Negafix. My book goes into detail about that for both colour and B&W, but I also published similar content on this website about scanning negatives before I undertook the the book project. www.luminous-landscape.com/tutorials/silverfast-scanning.shtml (http://www.luminous-landscape.com/tutorials/silverfast-scanning.shtml) and http://www.luminous-landscape.com/techniques/scanning_colour_negatives_raw_or_not.shtml (http://www.luminous-landscape.com/techniques/scanning_colour_negatives_raw_or_not.shtml). All this predates SilverFast 8 but the principles are pretty much the same.
For 35mm colour negs, if you can get your hands on a really good dedicated film-scanner such as a Nikon 5000 or similar, it will likely deliver a sharper result than the Epson V700. The V700 however is very good for the larger format negatives (for example 6*6 cm and larger).
I believe Dean, who has been participating in this thread, if I remember correctly has also been quite successful digitizing film using a camera instead of a scanner, so he may well pitch in here.
As always - choices, choices, choices.
-
Hi Chadd,
I should have mentioned - if you would like to see some samples of colour negatives I scanned in SilverFast using Negafix and converted to B&W, some with and some without application of effects (toning, colour accents), you can have a look here http://www.markdsegal.com/angkor/ (http://www.markdsegal.com/angkor/). The scanner used at the time was a Minolta Dimage Scan Elite 5400 (dedicated film scanner) - again one of those excellent pieces of scanning hardware that was discontinued. While you are on the site http://www.markdsegal.com/ (http://www.markdsegal.com/), the galleries of Vientiane and Bangkok-Wats are also from scans of colour negative film done in SilverFast with the same scanner.
-
One question though: does the profile replace the scanner profile in silverfast, or does it get applied to the image in photoshop after scanning?
Either way works. Although, if you go the route of assigning it yourself in photoshop you are cutting your workflow in Silverfast short at the point when you exported your target scan. May or may not be a disadvantage to you.
I didn't get a chance to play with CoCa again today, but I did collect a new watercolour pencil drawing to reproduce that I'll run with a CoCa profile. I'll post here about how it goes.
-
Either way works. Although, if you go the route of assigning it yourself in photoshop you are cutting your workflow in Silverfast short at the point when you exported your target scan. May or may not be a disadvantage to you.
Stephen yes, you're right - either way works, but just to amplify on what those ways are: two options here: (1) do a "raw" scan in SilverFast, then open the image in Photoshop and "Assign" your scanner profile. OR (2) In SilverFast Preferences> CMS select the scanner profile in the Input/scanner pane and make sure Colorsync is also active. As you are purposing the images to a print, I recommend scanning in ProPhoto colour space and using ProPhoto in Photoshop as well. And that means making sure to scan and work in 48-bit RGB (i.e. 16-bit per channel). Keep your master file in 16-bit, and if you need for example sRGB JPEGs for quick transmission, create copies in Photoshop that you shrink to 8-bit for this purpose.
-
I got a chance to run the job I collected yesterday through CoCa profiles. They are very useable for my purposes. Different to the PM5 generated profiles, but just as good, if not better in some respects. I have slightly fewer and simpler corrections to do to get a match than I do with the PM5 profiles.
The LAB Clut algorithm is the one that I use, whitepoint pushed to 1.2, quality Ultra High (might as well, doesn't take THAT long). Needs a slight exposure correction after assigning but at least the highlight data is safely in the masterfile.
The attached picture shows the difference between the algorithms. No white point tweaks or extensions on these. The crops are, from left to right, Gamma Matrix, LAB Clut, Shaper Matrix, XYZ Clut and PM5.
-
Stephen,
They all look pretty good to me and hard to see much difference between them. What I would find most useful, if you are positioned to do it, is to see how those results compare with using a profile for your scanner from SilverFast or another external provider, so we can compare the difference it makes relative to the careful procedures and high-end materials you are deploying for this purpose.
-
I don't know if CoCa qualifies as "high-end". CoCa is free. So, maybe the results are "high-end", but certainly not the price. ;)
-
Ya but the target he's using costs between USD 250~300 :-)
-
Colour is pretty similar across those images, but if you look into the yellow band you'll see that the second one, LAB clut algorithm, is the smoothest. The others are all a bit bandy, like the brightest bits are on the ragged edge of clipping.
I'll post similar tomorrow using the epson canned profile, an HCT profile and whatever other ones I can dig up. I'll try get CoCa to build profiles from my CC24 and qpcard203 if I have time too.
Hmm, I seem to have collected quite a few targets.
-
one other issue of note, whether it applies to this problem or not, is that some scanner profiles clip whites when converting to working space depending on intent. Monaco scanner profiles, for example, do so using perceptual, therefore requiring colormetric. THere is a lot of great info about scanning and color management here, if you dip deep-
http://www.hutchcolor.com/index.html
Tyler
-
one other issue of note, whether it applies to this problem or not, is that some scanner profiles clip whites when converting to working space depending on intent. Monaco scanner profiles, for example, do so using perceptual, therefore requiring colormetric.
Dang. Haven't played with rendering intent when converting from scanner profile to working space. Just assumed it was fixed at relative like when you convert between working spaces.
-
I'll post similar tomorrow using the epson canned profile, an HCT profile and whatever other ones I can dig up. I'll try get CoCa to build profiles from my CC24 and qpcard203 if I have time too.
OK so it won't be today. I've been playing with the demo of Silverfast a bit, its V700 profile seems pretty good. I'll put something together and post again soon.
Stephen
-
Looking foreward to what you come up with Stephen.
-
Hi Mark, yes, they are! I was able to get mine effectively unused off "e" bay ... for $120. If you're patient and very persistent you can often get lucky.
BTW, your Angkor gallery are really awesome! Thank you for sharing.
Stephen, I like the results your getting with Coca. It will be interesting to see how they compare to the canned profiles. Although I've finished with the stamp project for now, I'm going to give this a try.
I'm getting more and more scanning projects so all this information is quite useful.
-
Thanks Chadd, glad you enjoyed them.
-
Posting a bit later than I thought I would, apologies, but last week got a bit busy. I've put together a set of images for comparison. All are saved in AdobeRGB so it's probably best to view them in PS unless you trust your browser. All files were passed through LR for the downsample from original to the res you see here.
final colour.jpg - from my masterfile. This file prints a close match to the original drawing.
SilverFast No adjustments.jpg - straight out of Silverfast
PM5 DCSG No Extension.jpg - Vuescan Raw; assigned PM5 generated profile; DCSG target; no highlight extension
CoCa DCSG Lab Clut No Extension.jpg - Vuescan Raw; CoCa generated profile; DCSG target; Lab Clut algorithm; no highlight extension
Three more crops and discussion in next post.
-
PM5 HCT No Extension.jpg - as for DCSG above
PM5 HCT Extended.jpg - as above, but highlights extended significantly
CoCa QP203 Gammamatrix WP 1.2.jpg - CoCa generated profile from QP203; Gamma+Matrix algorithm; White Point 1.2
Haven't thought this through into an essay so here are some points that have struck me:
- Silverfast's generic V700 profile yields very good colour. It does compress the contrast, however, which would need to be fixed to get a match. Mainly it's the highlights that get squished - instead of being bright and clear they are a little murky.
- There really isn't much between the PM5 and the CoCa profiles. I can get a good match fairly easily with either. PM5 seems to retain a little more detail in the highlights, but loses a nitpicking touch of absolute accuracy. I'm running with the CoCa profiles for now and will continue to do so unless they start presenting issues.
- note how massively the un-extended HCT profile clips. not an appropriate target for this work. it's great for photographic prints, of course.
- the extended HCT profile enters the realm of useability, but needs lots of contrast and colour work for a match. (this is actually how I started this kind of work last year, before trying the DCSG)
- for interest I included a CoCa profile generated from my QP203. Generic reference file. The Lab Clut algorithm gave awful results, but the Gamma+Matrix with a white point tweak seems to be OK. It's rather punchy and oversaturated, but it's not a horrible profile. Not bad for a $50 target and not-ideal reference data.
That's it for now.
All of these crops are reproduced with the kind permission of the artist. If you are interested in seeing a bit more of the work, please visit her site lindsaysdesigns.com (http://lindsaysdesigns.com)
-
A sleeper! Attached is a scan from Epson Scan - the utility that comes with the scanner. I've never delved into it until now but I decided 10 minutes ago to try and get it to produce as flat a scan as it possibly could. I zeroed all the automatics, levels and curves that I could find, told it to use its generic scanner source profile and then convert to AdobeRGB.
It's incredibly similar to the Silverfast results! The silverfast scan has better colour, I think, but there's not much to it.
-
Hello Stephen,
First, we all owe you much thanks for taking the time and trouble to run these tests and prepare the results. I know from my own experience designing, performing, evaluating and writing-up tests, it takes time and concentration.
I have opened all your provided files in PSCS5, looked them over on my properly calibrated and profiled NEC PA271W display, and have the following observations, assuming that the "final" is the one that looks closest to the original media that you are scanning; I see them falling into three categories
(1) PM5 DCSG No Extension and PM5 HCT Extended look to me as the ones that provide the closest "out of the box" outcomes to the final.
(2) The two CoCa and the PM5 HCT No Extension look to me the riskiest in terms of losing detail to either saturation or highlight issues.
(3) The SilverFast and EpsonScan results are indeed very close to each other, but both showing less contrast and brightness than the "final". This also makes them safer than category (2), because it is very easy to add-in a bit of punch and brightness, totally under the operator's control, to get to "final" without risking loss of wanted image detail.
Hence based on your results from this sample, which is a useful one, I would be comfortable with the profiles in categories (1) and (3), using whichever is the most convenient and least-cost to obtain.
-
Stephen, your test is very interesting. Thanks for posting your results.
The profile that provides the best match with the original art obviously is the best profile for your purposes. The potential problem I see, however, especially in trying to extend your test results to film scans with lots of detail and subtle color and tone differences, is that most of your samples have very significant highlight clipping (except for the PM5 HCT Extended which just mainly clips the white area and the Epson and Silverfast profiles which have minor clipping). Just looking at your reduced sized jpegs, it appears that many of your profiles result in some loss of detail in the clipped areas. To me, from a detail perspective, it looks like the PM5 HCT Extended profile may be the best, followed closely by the Epson and then the Silverfast. It would be interesting to see if you could open the PM5 HCT Extended or Epson version in LR and make a few quick adjustments that would give you the color match similar to your final version, but with a little more detail.
The problem I find in determining if there is true highlight clipping in scans, especially in the individual color channels, is that no desktop scanning software that I know of permits output of true RAW data from the scanner CCD. So, it’s difficult to determine whether you have RAW clipping or the clipping is introduced at the profiling stage. Furthermore, clipping could be introduced, extended or hidden upon conversion from your custom profile to another color space such as aRGB. Depending on your scanner and scanning software, the best you may be able to do is check for clipping at multiple stages to see what you can learn. For example, if you can output 16 bit linear data, you can check that data. You can check for clipping upon applying your custom profile and again after you convert to aRGB or another color space.
Of course not all clipping is necessarily harmful. Also, if you can’t physically alter the amount of light reaching your scanner’s CCD, then you may have no way to prevent RAW clipping. And the most important caveat – if you’re getting the results you want despite some clipping, then ignore everything I’ve written about clipping! :-[
-
Dean, I'm wondering firstly, whether we're downloading and looking at the same stuff. In my post above I was examining the JPEGs that Stephen attached as illustrated in his posts. Did I miss anything Stephen?
Secondly, assuming we looking at the same stuff, I don't get the extent of highlight clipping you are referring to except in the category (2) scans and to a minor extent only a small bit of red channel clipping in the category (1) scans. Also recall these are compressed JPEGs, so some allowance needs to be made for the possibility that data loss may be adversely affecting our appreciation of this issue. The Silverfast and Epson scans are not clipped at all in my version of PSCS5. I am opening the files using the "Preserve Embedded Profile" option. Are you doing the same? I find the Epson and SilverFast results to be pretty much interchangeable and unclipped, which would help explain why they are of somewhat lower contrast and brightness - and therefore "safer" starting points.
Thirdly, my understanding is that these are reflective scans, which should have been made from profiles generated with reflective targets, and if that's the case, until it's tested it's not clear what they imply for any particular profile generated for transparencies from a transparency film target.
In both SilverFast and Vuescan, we can output a linear gamma scan with no further luminance and colour adjustments, which is the closest we'll get to producing what the CCD registers. So we can assess clipping on that basis as a starting point. A question for Stephen would be the gamma factor used in his profiling and scans, and whether it is the same across all samples.
-
Mark, I downloaded the jpegs, opened them from Bridge in PS-CS4 in layers with the embedded aRGB profiles, then added a levels layer on top to check for clipping. My method shows significant highlight clipping, particularly in the red channel, in 5 out of the 8 samples. The other three samples also show clipping as I described above. I have no idea why you get different results.
Yes, the jpeg compression could account for some or all of the clipping, which is why I mentioned that I just looked at Stephen’s reduced sized jpegs. Conversion from Stephen’s profiles to aRGB may have also contributed to the clipping. These are issues that Stephen will have to look into as I described in my prior post.
It’s difficult for me to make a fair comparison of the Epson and Silverfast profiles due to their different sizes. They look very similar, but I think I detect a small difference. I could be wrong.
Yes, Stephen’s scans are reflective scans, and I’m not sure how much can be extrapolated to film scans. My point was about clipping and whether some profiling methods introduce clipping. If so, it would be something to look for when using CoCa, for example, for film scans.
I use 16 bit linear output as one way to check for clipping (highlight and shadow). It’s not as good as RAW data, but, as you mentioned, it’s the closest to RAW as we can get. In other words, there is a possibility that however you examine a 16 bit linear tiff, the fact that it shows or doesn’t show clipping is no guarantee that the RAW data does or does not have clipping. Since I have no way to look at the RAW data, I’ve tried to convince myself that the 16 bit linear data is close enough. ::)
-
Thank you all for helping me challenge my thinking and workflow. Too easy sometimes to just plod on with something that works, but may not be the best approach.
OK, refining things a bit: here are the tiffs (https://dl.dropbox.com/u/18169539/LuLa%20scan%20thread.rar) from which I created the jpegs above. Should have done this in the first place. Except for the linear source file they are all in ProStarRGB, which is another thing I'm playing with at the moment. In this situation it's just a container - I've performed no adjustments after conversion from the scanner profile so it should not be influencing these results.
The LR downsample+jpeg+AdobeRGB process contributed to the clipping and obscured the actual results I got. I also capture sharpened the files, 0.6 87 Smart sharpen, which probably added a little clipping too as I did not protect the highlights. lazy.
Yes I do have clipping at the upper limits. No it's not clipping that I'm worried about as the files print fine. It's mostly in the paper colour background, touching on the brighter colours in the drawing. The approach I used in the end was what similar I described a few posts above: CoCa DCSG profile extended by setting the whitepoint at 1.1. This gave me a slightly compressed masterfile that I then brought back to the exposure I wanted, while trying to avoid clipping. It's a controlled approach: I can apply as much compression/extension as is necessary to capture a given original. The Silverfast output is definitely safe, but will create unnecessary correction work if the gamut of an original fits well within that of my target.
A gap in my understanding: I'm not sure what you mean, Mark, by 'the gamma factor used in my profiling and scans'. I'm not applying any gamma adjustments to my scans. I'm relying on the profile to restore correct exposure to the linear source file. I output 16-bit linear, assign the profile, convert to working space and the exposure and colour are 95% of the way there. The Silverfast and Epson scans have been gamma adjusted, but out of my control.
I tried what Dean suggested: use LR to bring the HCT Extended sample close to the look of the 'final colour' sample. It came up nicely, retaining good variation in the oranges that is a little lacking the CoCa-based final. I suppose the measure of a good profile is not just how close it gets you when you assign it, but how well the resultant file responds to adjustments. How robust it is. HCT-based profiles yield very robust files, for reflective and film work.
-
Stephen, I may have been a bit cryptic in that one phrase - sorry. Of course a profile describes device behaviour, so that isn't where the gamma issue comes in. It comes in at the image interpretation stage - whether the default interpretation has gamma = 1.0 (linear) or some higher gamma which increases brightness and contrast, and along with that higher risk of clipping. The fact that you output 16-bit linear is great. SilverFast by the way does allow you to control the gamma in Preferences. The profile wouldn't have much impact on how well the file responds to adjustments unless it is so bad that the ensuing corrections could exceed the capabilities of the image editing software. Bit depth and gamma would have a bigger impact on how well the file responds to adjustments.
-
OK gotcha now. You were referring to the brightness/contrast adjustments made to a file after applying the scanner profile, yes? If so then in most of my samples there have been no adjustments. Only the 'final colour' sample had a (slight) exposure lift to bring it up to match the original. The others were just assign, convert to working space and save, with Epson Scan and Silverfast taking care of these steps for me.
'quality' of profile makes a big impact, particularly when scanning film. Slides are often underexposed and being able to restore them and maintain good colour is very important. When scanning bright artworks it's not such a big deal. You don't need to push exposure up by 1-1.5 stops, small tweaks are all that's needed. The are examples of this on the HCT site, but I'm going to try and make my own, if only to justify the purchase of my HCT transmissive target. :-)
-
'quality' of profile makes a big impact, particularly when scanning film. Slides are often underexposed and being able to restore them and maintain good colour is very important.
Stephen, As you may know, I've been doing this for years and a whole chapter of my book (http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml) is devoted to the subject of colour managing scans and post-scan workflows, including the importance of correct profiling. Yes, if you use the wrong profile or a really crap profile it can make a big difference to the range of risks, time and trouble encountered in a post-scan workflow, but as long as the profile is the correct one for the scanner model and made with reasonably decent profiling software so that neither blacks nor whites are clipped, there are a number of effective techniques that will reveal underexposed detail that exists in the original media, often to surprising effectiveness, given how dismal the original media may appear to be. I have also dealt with that subject in my scanning articles here on LULA and in the book. Of course I agree the better the profile the less work this involves, but my experience is that no matter how *accurate* the profile, these heavily under-exposed areas do need a fair bit of user intervention regardless. So yes, a decent profile is necessary, but not sufficient.
-
no matter how *accurate* the profile, these heavily under-exposed areas do need a fair bit of user intervention regardless. So yes, a decent profile is necessary, but not sufficient.
I'm going to test how 'decent' the profiles I have available to me are, with respect to restoring underexposure. For reflective I've got profiles from HCT, DCSG, Epson, Silverfast, CC24 and QPcard. For transmissive I've got profiles from HCT, Epson and Silverfast.
I don't when I'll get to this, or how exhaustively I'll do this, but it'll be fun and I'll bring the results here (probably a new thread) when I've got some.
-
That's great Stephen, it will make a useful contribution.