Pages: 1 [2]   Go Down

Author Topic: NDQ - Scanning Resolution  (Read 16441 times)

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
NDQ - Scanning Resolution
« Reply #20 on: February 26, 2006, 09:14:17 pm »

Mark,
I'm still a little puzzled. You seem to have done a lot of testing and comparisons of different scanning software and procedures. Have you actually compared a scan at full optical resolution using the software's built-in ICE and/or grain reduction technology (or neat Image), with one scanned at say 3100 dpi and later processed with Neat Image?

It seems to be a fact that noise and grain cannot be eliminated without some loss of resolution, however small. It's usually a compromise between minimal resolution loss and maximum grain reduction that one aims for.

It seems to me reasonable that there might be an advantage to final quality if grain reduction is applied to the full size image (whether in Neat Image or the scanner software). When such an image is subsequently downsampled for printing, grain will also be further reduced.
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
NDQ - Scanning Resolution
« Reply #21 on: February 27, 2006, 01:35:35 am »

Ray,

I print up to A3. The quality of detail I get on A3 prints is outstanding without scanning at 5400 DPI, so I don't do it for reasons I've explained above, unless the combination of a crop and the intended print size takes me there.

I only use ICE on images that need massive sludge removal. It slows down scanning alot, and when you have many negatives to scan the productivity loss is substantial. It is great for dirt and specks but nothing special for grain. Neat Image is far superior to the scanner software when it comes to fine control over grain removal and the results thereof. As for the role of downsampling in grain reduction - maybe it happens as a by-product of resolution loss, but I think it preferable to deal with the balance between grain reduction and sharpening in a direct, deliberate and controlled manner with purpose-dedicated tools, which downsampling doesn't give us.

Scanning at 5400, using ICE, using scanner grain removal software, and downsampling adds up to a workflow that I haven't tested because the individual components seem to me somewhat sub-optimal for the purpose at hand to me. My scanning got interrupted for the past several months as a result of a very large digital imaging job I just completed, but when I get back to scanning I may nonetheless try what you are proposing just out of curiosity to see how it compares.
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
NDQ - Scanning Resolution
« Reply #22 on: February 27, 2006, 12:20:55 pm »

Ray, I'd like to revert to my last post replying to your lingering puzzlement. I actually have done some tests at full optical resolution, though not exactly in the workflow sequence you mention. The work I refer to is a collaborative exercise between Harald Johnson and me on the whole issue of optimal scanning resolution, and it is published in his "Mastering Digital Printing Second Edition", pages 83 to 85, published last year by Thomson Course Technology, Boston MA.

The section is preceded with relevant conceptual material written by Harald from page 78 onward. In writing this material, Harald consulted with Wayne Fulton (author of "A Few Scanning Tips" www.scantips.com, a very understated title relative to what Wayne's book contains).

I would just like to take a very few extracts from Harald, to help you better understand where all this is coming from.

On selection of scanning resolution, Harald reports Fulton's advice: You scan for the capability of your output device, which means that you choose a scan resolution based on the needs of the monitor or printer that will display the image in its finished state. Harald goes on to say that there is no one answer and varied opinions about optimal scanning resolution.

That is why we did the tests, looking at maximum optical resolution, even integer resolution, odd integer resolution, and intentionally low resolution. We selected a photo I made of the Cairo Tower using ASA 100 colour negative film, because its details, highlights, shadows and diagonals lend themselves well to observing the consequences of different resolution settings. I performed all the scans and made all the prints strictly according to the required parameters of the exercise pre-discussed between us, and sent Harald all the results on disc and on paper. We standardized the output at 6.7 by 10 inches of printed image and adjusted input and output resolution settings to maintain the same print size. We started at 5400 PPI and worked down to 860 input/120 output, taking care to scan for integer and non-integer divisors along the way. No other image editing was done.

We both examined all this material carefully and came to the same conclusions, which are illustrated and reported in the book. They are that (1) degradation clearly visible with the naked eye set in below 1704 input/240 output, with the biggest hit at 860/120. (2) It required a 6X loupe to see the very slight difference that using integer divisors makes, and in general (3) if you don't look at prints with a powerful magnifier, there is a wide latitude of scanner settings to produce similar quality results.

By the way, Harald's book has been well-recommended on this website.

Hope that helps.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Andrew Teakle

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 90
    • http://www.andrewteakle.com
NDQ - Scanning Resolution
« Reply #23 on: February 27, 2006, 05:23:31 pm »

Mark,

Thank you for your very informative article. You have explained your rationale for choosing your scanning resolution very well and it obviously works well for you.

When we return from a a trip (or returned, because we're mostly completely digital now, with just a Mamiya RZ67, and 35mm for star trails...left in the film realm) we're not sure of the intended output of our images. Our images are represented in corporate galleries and clients will request them at varying sizes depending on their needs. We then have them printed and the gallery provides framing etc. We therefore have to scan at the maximum optical resolution and then only perform all the image optimisation once. While the scan times are longer (we also use ICE routinely) the rest of the workflow is not a lot slower than for a smaller scan. We can then resize as necessary when we get an order, but it is always ready to go if we get an order fo a very large print. This is also very important when we get an order while we're travelling and don't have access to the scanner to make a one-off maximum scan.

Of course the whole process takes more time and more storage, but works best for us in the long run.

Thanks again for the article,

Andrew
Logged
Andrew

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
NDQ - Scanning Resolution
« Reply #24 on: February 27, 2006, 05:45:18 pm »

Andrew,

Glad you enjoyed the article. Of course what you are saying makes perfectly good sense for your situation - when you can't know ahead of time what kind of request is coming your way, best is to scan and process the image once, keep the most information you can in a central file that never gets resampled, and then preferably re-scale a copy, or if you really need to - resample a copy for individual orders as the case requires.

You say you do ALL the image optimization at once. If you are using PK Sharpener Pro, of course you would know that Output Sharpening is resolution and size specific, so you would do that step custom for each order as the last thing before printing, and discard those sharpening layers before re-saving the file.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
NDQ - Scanning Resolution
« Reply #25 on: February 27, 2006, 08:42:31 pm »

I'm a bit concerned about blanket statements such as, 'whatever works best for you...' or, 'different strokes for different folks', or 'this person who wrote this book is highly regarded and (implication) must be right" etc, etc.

I'd really prefer some clear principles and reasons. If I can summarise what I've learned from this thread so far, it is:

(1) There is no time saved at the scanning stage when scanning at less than full optical resolution, however,

(2) There is a saving in time at the processing stage which might include running the file through Neat Image as well as manipulation in Photoshop. It's always faster to work with smaller files.

(3) There is also money saved at the archival stage for smaller files. They take up less hard drive space and/or fewer DVDs and less time to record to DVD.

There can, therefore, be good economic reasons for scanning to output print size, especially when there is no reasonable expectation that larger prints might be required at a later date.

In favour of scanning at the full optical resolution of the scanner, we can summarise the following points.

(1) There is no increase in scanning time; a point worth repeating because the slow scanning time of desktop prosumer scanners has always been a big issue and the misguided expectation that time can be saved by scanning at a lower resolution is very seductive.

(2) Scanning at full resolution definitely allows for the possibility of extracting more detail from the slide or negative, provided the detail is there to begin with, and at this point it should be mentioned that the resolving capacity of a 5400 dpi scanner is not as high as you think. Just as with lenses and camera sensors, the maximum resolution will always be worse than the theoretical resolution limit of either the lens or the sensor. For example, a 5400 dpi scanner should be capable of caturing 100 lp/mm. A fine grain B&W film with a first class prime lens used with impeccable technique, spot on focussing, rock solid tripod, stationary subject etc should also be capable of capturing 100 lp/mm. Unfortunately, capturing such fine detail with a 5400 dpi scanner is not going to happen due to further MTF loss from the scanner's lens, and no doubt other factors.

(3) If the purpose of scanning is to preserve the information in old, continuously deteriorating film, as well as make prints, then it's really a no-brainer. Scanning at full optical resolution is the only sensible thing to do.

Lastly, the issue of a possible quality advantage visible in largish prints (A3 or A3+) in the context of extensive processing of the full rez image before downsampling, is of concern to those of us interested in art-type prints and interested in extracting the finest quality from the film, whatever the print size.

Mark, I look forward to seeing your conclusions regarding this last issue. One concern I have with SilverFast is its lack of GEM, which Minolta includes in its software with ICE. Silverfast has its own grain reduction technology which is much more elaborate and complex in its controls, but not necessarily ultimately better or even as effective as GEM. The next time I do some scanning I'd like to compare a 5400 dpi scan with GEM turned on with a 3100 dpi scan processed in Neat Image.

Cheers!  
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
NDQ - Scanning Resolution
« Reply #26 on: February 27, 2006, 10:15:34 pm »

Ray, re your second (2) and (3), whether the resolution limit in PRACTICE is due to the lens or the sensor, those are baked into the product we buy, and once bought the remaining issue over which we have control is the choice of resolution. In the final analysis the resolution setting that gives us the most resolution we can see on the largest print we will make from that negative without a loupe is the sensible setting. That is what practical testing is aimed at determining and that is the context of any testing I have done. I'm focused on visible differences between one setting and another, one workflow practice and another, and practical efficiency and economy in working. If this takes me to full optical resolution for certain print sizes or magnification ratios that I'll be using, then full optical resolution would be correct.

Re your point immediately below second (3), let us examine the A3 situation for a moment. If we want the longest dimension of our A3 print to be say 14.5 inches (on a 16.5 inch A3 sheet), and we want 360 PPI of output resolution, then the scanning resolution will be 5220 PPI, which is so close to 5400 it isn't worth testing. So this only makes sense to play around with if we are satisfied with something meaningfully less than 360 PPI - say 240 PPI of output resolution; then we could get away with scanning at 3480 PPI. So at 240 for example, the issue boils down to whether for an A3 print we would see a difference in image detail by (i) scanning at 3480 PPI, or (ii) scanning at 5400 PPI which yields 372 PPI output resolution, and downsampling this to 240.  I don't mind testing this even though based on what I've already done I think the answer is predictable; but perhaps you should test it too. We can then talk about our respective results even though we will not have standardized either the image or the quality criteria or seen eachother's prints, because I think we live half a world apart from eachother, so very "impressionistic" stuff, but anyhow not necessarily useless.

As for grain removal software (your last para above), I've tried all that stuff. Give me Neat Image any day. It's faster, more controllable and more effective; what's more it isn't necessarily baked into the image, as I can apply it to a duplicate image layer giving me another universe of options for controlling and targeting it.
« Last Edit: February 27, 2006, 10:17:12 pm by MarkDS »
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
NDQ - Scanning Resolution
« Reply #27 on: February 28, 2006, 09:38:58 am »

Quote
Re your point immediately below second (3), let us examine the A3 situation for a moment. If we want the longest dimension of our A3 print to be say 14.5 inches (on a 16.5 inch A3 sheet), and we want 360 PPI of output resolution, then the scanning resolution will be 5220 PPI, which is so close to 5400 it isn't worth testing. So this only makes sense to play around with if we are satisfied with something meaningfully less than 360 PPI - say 240 PPI of output resolution; then we could get away with scanning at 3480 PPI. So at 240 for example, the issue boils down to whether for an A3 print we would see a difference in image detail by (i) scanning at 3480 PPI, or (ii) scanning at 5400 PPI which yields 372 PPI output resolution, and downsampling this to 240.  [a href=\"index.php?act=findpost&pid=59177\"][{POST_SNAPBACK}][/a]

Mark,
You're fudging the issue  .  With large prints, 360 ppi is quite unnecessary on Epson printers. According to my maths, a 5400 dpi scan of 24x36mm film produces a 22"x33" print at 240 ppi with no interpolation.

Perhaps another reason I find no incentive to scan at a lower resolution is because I use Qimage which automatically resizes and resamples an image before sending it to the printer. I could open a 120MB file in Qimage that is just 1"x1.5" and 5400 ppi, and whatever print size I specify (6"x4" or 22"x33"), Qimage will do whatever resampling is required before printing, including final sharpening.  

Nevertheless, I am prepared to do some tests on this issue when I find the time   .
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
NDQ - Scanning Resolution
« Reply #28 on: February 28, 2006, 10:02:42 am »

Hi Ray - no I'm not fudging anything. I agree with you that 240 is fine for large prints. BUT there are knowledgeable practitioners who for whatever the reasons prefer to print at 360, so I was simply making the point that what one does with the input partly depends on what one wants as output. That's just good cooking, but not fudge.  

I think I'm back to some scanning this week, and if so I'll try the option at 240, where there is scope perhaps to see any differences of outcomes between input resolution settings. I'll post a comment on the results here.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

tkk

  • Newbie
  • *
  • Offline Offline
  • Posts: 3
NDQ - Scanning Resolution
« Reply #29 on: February 28, 2006, 06:13:00 pm »

I echo Ray's remarks. The procedures is complicated but understandable due to the limitations of the equipment - principally the Minolta scanner. This scanner has technical imperfections. The values of both 16 bit and 5400 bpi is questionable. I have a Nikon 8000ED scanner and I has been able to make NDQ, even EDQ quality outputs for the finest quality large prints quite easily using the following simple steps:

1) Scan at the max 4000dpi for 35mm and MF. There is no need to do calculations to figure out scan resolutions for the print resolution. That adjustment can be done later in Photoshop with much better result.
2) Turn on NikonScan Digital GEM feature, which reduces or eliminate grains as good as Neat Image Pro without all the fuss.
3) If an image has significant shadow areas or underexposed, scan at 14 bit. Otherwise 8 bit is more than good enough for all except the most demanding fine art prints.

Film grains, even for the finest grain films, make scanning at 4500 or higher resolution a waste. The build-in Digital GEM software of Nikon does as good a job as Neat Image Pro without spending the money. 16-bit scanning far exceeds the dynamic range of films and is a waste or both time and file size. Also, Nikon scan software is so good that it can produce highly accurate image of a large variety of negative films. (I have measured them using a photo spectrometer.) There is no need to pay extra for Silverfast or learn to use it. (I use Silverfast, however, for my flatbed scanner.) Finally, post-scan sharpening is best done in Photoshop using a variety of tools and methods that can be found in books and such.  

In short, Mark's complicated and expensive procedure is necessary only for those who unfortunately owns the Minolta Dimage Scan Elite 5400 scanner.  
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
NDQ - Scanning Resolution
« Reply #30 on: February 28, 2006, 07:21:05 pm »

Hello tkk,

The Minolta scanner won design awards when it was produced and has received excellent reviews. Many people including serious professionals use them successfully. I find the detailed image quality it produces remarkable. Of course there are always people who have negative experience with any piece of equipment, but that is not necessarily a reflection of overall quality. I'm sure Nikon also makes excellent scanners, and there are probably people who complain about them too. However, all that said and done, the workflow I described has a logic of its own beyond the make of scanner one uses.

Any high resolution scanner will reproduce film grain. You can always use GEM but it bakes the results into the scan, and if you are happy with that, you don't need Neat Image. I'm not. As I said in a previous post and in my article, using Neat Image in Photoshop after scanning - especially on a layer of its own, regardless of the scanner, gives you avenues of control over grain management you don't otherwise have.

The option of choosing a scan resolution at scanning time rather than resampling in Photoshop again is one of those choices that is scanner-independent. One can spend a month of Sundays debating which option ends-up producing better quality under what circumstances. Ray and I have been having that discussion in this thread, and the up-shot is that we may be doing some comparison scans. Should be fun. Have you tested both options and compared the results? I'd be interested in your observations from actual tests, and how you specified the tests.

For a great many situations the advantage of having 5400 PPI rather than 4000 PPI to play around with is nil - as you say, but those extra PPI can come in handy for very large prints, or large prints from cropped images, etc.

I doubt one would see any quality difference under most circumstances whether   scanning at 14 bit or 16 bit. Scanning at 8 bit is another matter. Again, alot of debate, but MOST experts in these matters overwhelmingly recommend using greater than 8 (16 if available) as the best insurance against banding and posterization, especially when working in large colour spaces and making substantial image adjustments in Photoshop.

Let us not confuse bit depth with dynamic range. This is explained here: http://www.scantips.com/basics14.html. Anyhow, I do make "fine art prints", and many images of all kinds have shadow areas for which a combination of high bit depth and high DR can only be helpful. So I vote for 16 bits, thank you.  

If Nikon's native scanning software is so good you don't need Silverfast, that is great, because Lasersoft charges a big bundle of money to provide this program for Nikon scanners. There must be a market and some reasons, otherwise they wouldn't waste their time on it, but for anyone who can do well without it - that's to their advantage.

Finally, the procedure I described in my article takes a certain number of words to explain, but honestly when you do it, really isn't complicated and expensive. Many professionals own Neat Image and PK Sharpener Pro anyhow, and if they don't, these are not nearly the most expensive plug-ins, nor are they difficult or time-consuming to use, after gaining a bit of experience. All I've described in that article can be boiled down to grain and sharpness management through two plug-ins, and perhaps some Photoshop layer masking in certain images where one size does not fit all - one has that luxury using the procedure I described. Many other procedures I have no doubt work just fine (that is the beauty of digital imaging, lots of options) - but not all procedures offer the same kind of control.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
NDQ - Scanning Resolution
« Reply #31 on: February 28, 2006, 08:55:01 pm »

Quote
I echo Ray's remarks. The procedures is complicated but understandable due to the limitations of the equipment - principally the Minolta scanner. This scanner has technical imperfections. The values of both 16 bit and 5400 bpi is questionable. I have a Nikon 8000ED scanner and I has been able to make NDQ, even EDQ quality outputs for the finest quality large prints quite easily using the following simple steps:

[a href=\"index.php?act=findpost&pid=59242\"][{POST_SNAPBACK}][/a]

tkk,
Thanks for your support, but that's not quite what I thought I was saying   .

When I bought the Nikon 8000ED, I thought to myself, 'this is probably going to be the last scanner I'll ever buy'. I intended to use it to archive all my 35mm film, as well as MF film that I was then shooting with second hand equipment that had become cheap because so many professional were switching to digital cameras.

Years later when the Minolta 5400 arrived, I didn't show much interest because I figured the 4000 dpi I already had would be enough for 35mm film. However, I still hadn't made much progress archiving old material and in the meantime I had switched to digital cameras.

The roll-out of the Elite 5400 ll happened to coincide with a renewed interest in my archiving my old 35mm film, so after some googling, reading a few reviews and test reports as well as engaging in a few discussions with Mark, I figured the Elite 5400 ll really would be capable of extracting more detail from 35mm film than the Nikon 8000ED. The price of the hardware was not much more than the cost of just the Silverfast software alone for the Nikon 8000ED, whch I was contemplating buying as an alternative to a new scanner.

There's no doubt in my mind that the 5400 really is capable of capturing finer detail than the 8000ED. It might not be so obvious in the sense that faint detail that is visible on the 5400 scan might not be visible at all on the 8000ED scan. More usually it's the case that faint detail visible on both scans is simply stronger, more contrasty and more prominent on the 5400 scan.

In general, I find the Scan Elite 5400 ll, in combination with Silverfast, preferable to The Nikon 8000ED with NikonScan software. I try to do as much adjustment and processing in SilverFast as possible, bypassing as much as possible any need for further processing in Photoshop.

Ideally, I would like to be able to do all grain and noise removal in Silverfast too, but there's something to be said for Mark's position of not embedding a noise reduction process in the original scan which might not be the best result one can get at present or in the future.
« Last Edit: February 28, 2006, 08:58:10 pm by Ray »
Logged

Andrew Teakle

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 90
    • http://www.andrewteakle.com
NDQ - Scanning Resolution
« Reply #32 on: February 28, 2006, 10:31:38 pm »

Quote
You say you do ALL the image optimization at once. If you are using PK Sharpener Pro, of course you would know that Output Sharpening is resolution and size specific, so you would do that step custom for each order as the last thing before printing, and discard those sharpening layers before re-saving the file.
[a href=\"index.php?act=findpost&pid=59166\"][{POST_SNAPBACK}][/a]
Mark,

You're right, of course. We do our output sharpening using PKSharpener after resizing the image and send this file to the lab. Just a small and easy step, and one that can be done in our mobile office (a 4WD Mitsubishi van and laptop with DVD burner. Ain't modern technology a marvel  )

Andrew
Logged
Andrew

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
NDQ - Scanning Resolution
« Reply #33 on: February 28, 2006, 10:48:43 pm »

Yes Andrew, it is a marvel. When you think of the technology going into the manufacture of these fine quality sensors in the cameras and scanners - the level of detail they can scrape, and then every time a printer comes out of my Epson 4800 with all those dots laid down just like you want them to be with stunning colour, it really is quite incredible, but true. We're generating seriously good stuff with ease and virtuousity that was unattainable only a few years ago.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."
Pages: 1 [2]   Go Up