Pages: [1] 2   Go Down

Author Topic: NDQ - Scanning Resolution  (Read 16434 times)

Ubtree

  • Newbie
  • *
  • Offline Offline
  • Posts: 7
NDQ - Scanning Resolution
« on: February 24, 2006, 04:41:55 am »

Mark, thanks for a great article.  I have had a film scanner for some time, and found it difficult to get started because of uncertainty about an appropriate workflow.  Your article provides the practical advice that I have been lacking.

I have one question.  You say that "there is broad agreement that output PPI should be in the range of 240 to 480 or less".  Is the upper limit specified in order to avoid unnecessarily large files?   If you are not worried about file size, is there any reason not to scan at the highest resolution available to you?
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
NDQ - Scanning Resolution
« Reply #1 on: February 24, 2006, 10:55:52 am »

First, thanks - I'm glad you find the article useful - that makes it worthwhile writing them.

On scanning resolution - yes apart from file size there are reasons not to scan for more than the maximum resolution you may need up to 480 OUTPUT PPI: (1) no point exceeding what you can actually see on a print; with today's inkject printers, Fraser and Blatner (Real World Photoshop CS2) advise that sending more than 480 OUTPUT PPI to the printer won't achieve anything you can see, (2) the more you send the longer the file handling and processing times, (3) at some point sending too much data to the printer can obscure detail rather than reveal it (Fraser & Blatner page 110). They advise working within 240~480. On large prints I have worked down to 180 or a bit less satisfactorily.

Don't fear to plunge in with your scanner and try stuff. What can happen - at worst you waste some time, but that is how we all learn. Take several images that have some clear diagonals, good range of luminosity (light to dark tones) and fine detail in them. Scan them at different resolutions and print them. Then examine the prints and see for yourself under your operating conditions what works best for you. Again, all you consume is a bit of time, ink and paper. No harm done and you've learned something valuable.

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

Ubtree

  • Newbie
  • *
  • Offline Offline
  • Posts: 7
NDQ - Scanning Resolution
« Reply #2 on: February 24, 2006, 01:59:32 pm »

Thanks for the clarification.  That's a great help.

As you suggest, I'll get stuck in now and see what progress I can make!
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
NDQ - Scanning Resolution
« Reply #3 on: February 24, 2006, 08:51:27 pm »

Great, let us know how it goes and what you find.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
NDQ - Scanning Resolution
« Reply #4 on: February 25, 2006, 03:33:57 am »

Quote
It is important to know the approximate output sizes before scanning, in order to set the scanning resolution appropriately.


What are you actually saying here, Mark? I recall years ago doing a comparison between  two A4 size prints, one produced from an image scanned at an approprite resolution for an A4 print (240ppi or 360ppi, can't remember which) and one scanned at the full optical resolution of the scanner in relation to the size of the film. The larger file was down-rezzed to the appropriate size and ppi in PS before printing. The smaller file was already at the right size and ppi. I couldn't see any difference between the two prints. But I bet I would have seen a difference if I'd later decided to up-rez the smaller file to make a bigger print.

The only advantage I can see in scanning at the resolution required for a current print is in a commercial environment where you are scanning for paying customers and time is money. If you are scanning your own film for archival purposes or to produce your own 'art' prints, I see only disadvantages in scanning at less than the scanner's full optical resolution. Is there something I've missed here?
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
NDQ - Scanning Resolution
« Reply #5 on: February 25, 2006, 09:29:31 am »

Hi Ray - yes, there is something you've missed. Please go back three posts to my reply to Ubtree, where I have explained the possible downsides of excessive resolution. There may be some opinion that there is no such thing as too much resolution, but on careful reading of much that has been written about this subject, those views seem to be not correct.

As a practical matter, with a 5400 PPI scanner, max resolution on a full frame 35mm image generates absolutely huge files in 16 bit mode. I'm already into the 60~70 MB range scanning at around 3100. Why waste time, storage space and risk sub-optimal resolution due to the mathematics of stuffing hugely excessive amounts of data through the printer? I guess it boils down to what the maximum resolution of your scanner is and how much cropping you do. If your scanner doesn't go much above 3000 and you are likely to print A3, then there is a good case for maxxing-out, but not when your scanner can do way more. As well, I think it is good practice to study the image before scanning it, to know the purpose and visualize what the final product should look like; that helps one to make all the sensible decisions through the workflow.

I agree that re-sampling downward and comparing A4 prints you won't see much if any difference from a variety of approaches, but that isn't much of a test  because down-sampling throws away data which is usually less challenging to image quality than up-sampling which invents data. Also A4 is a forgiving print size. You may have observed that when you print to A3 or larger quality differences can become more visible.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
NDQ - Scanning Resolution
« Reply #6 on: February 25, 2006, 10:11:43 am »

Quote
As a practical matter, with a 5400 PPI scanner, max resolution on a full frame 35mm image generates absolutely huge files in 16 bit mode. I'm already into the 60~70 MB range scanning at around 3100. Why waste time, storage space and risk sub-optimal resolution due to the mathematics of stuffing hugely excessive amounts of data through the printer?


Hi Mark,
I still don't get it. The scan is the digital negative that replaces the chemical negative. Why would you want to make a lower resolution negative after going to all that trouble massaging the data in Silverfast, sometimes a fairly time consuming process as you know? Having too much data is no problem nowadays with cheap memory, DVDs and hard drive space and it's getting better as time goes by.

I understand the principle that there is no point in sending unnecessary amounts of data to the printer. There's no problem with resizing and down-rezzing is there? Why restrict your options?
Logged

dmerger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 680
NDQ - Scanning Resolution
« Reply #7 on: February 25, 2006, 01:43:49 pm »

Like Mark, I use the Minolta 5400, Neat Image, Photokit Sharpener and Photoshop CS, but I use the Minolta software and so far I've just scanned Velvia 50 and some Provia 100f.  I have a lot of negative film to scan one day, but I haven't gotten around to my negatives, yet. So, my experience is not identical to Mark's but is fairly similar.

I always scan at 5400 ppi.  I understand that if I scan at something like 3100 ppi, what actually happens is my Minolta will actually scan at 5400 ppi and then just resize down to 3100 ppi.  

I prefer to scan once, process my photos with Neat Image, PK Sharpener and Photoshop and then resize and sharpen a copy for printing.  All my editing is on layers and nondestructive.  The entire process is very time consuming.  I want to do it only once for each photo, so I prefer to work with and save the highest resolution, best file that I can make with my Minolta.  

My files are large, but I manage my work flow so that file size is not a significant problem and storage space is cheap so I'm not concerned about storage.

Also, my experience with Neat Image is different than Mark's.  I tried profiling each photo and creating my own Velvia profile with Neat Image using various photos.  My results were good.  I then created a Neat Image profile using an IT-8 target and my results were nearly perfect.  I now get a near total elimination of grain and no loss of fine detail.  

Perhaps Mark's experience with Neat Image profiles is due to his scanning at about 3100 ppi. If his scans are actually automatically resized down from 5400 ppi to 3100 ppi, then perhaps the resizing alters the grain structure in slightly different ways for each photo, which prevents him from creating and applying a standard Neat Image profile.  

In any event, there is not just one, correct way to scan.  My method works for me.  I'm sure that Mark's method works for him.
Logged
Dean Erger

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
NDQ - Scanning Resolution
« Reply #8 on: February 25, 2006, 03:06:54 pm »

Ray, we're talking past eachother. As I've mentioned above, there is expert opinion that large-scale down-sampling can cause loss of resolution.

As for the rest of it, each to his/er own. Anyone who cares less about time and storage space can by all means scan everything at 5400 PPI if they feel more secure about it. I'm the type who likes to get things done efficiently, and I've tested for what produces prints that are professional quality.

Cheers,

Mark
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 #9 on: February 25, 2006, 03:30:25 pm »

dmerger, the question of what the scanner does at various resolution settings is not new and I've verified this item. Minolta's technical people here in Toronto (while they were still in business - all gone now, unfortunately) told me that the scanner does not behave the way you suggested. As well, you may have noticed the substantial difference in scanning time that occurs as a function of resolution. That should give you a hint the difference is not simple down-sizing, which is a software calculation that happens in a blink of the eyelid.

As I mentioned in my article - or perhaps the Silverfast installment to come, one of the unique differences between negatives and chromes is that there is no IT-8 profiling for negatives, hence the kind of standardization that may be feasible in this respect for chromes is not available for negatives. That is why Lasersoft Imaging developed Negafix in Silverfast.

In any case, profiling deals with colour balance and luminosity, not grain. As well, grain structures of chrome and negative films are completely different - the film structure, manufacturing and development processes are very different. In either case, however, higher scanner resolution would tend to emphasize grain more than lower resolution. But since the scanner does not resample in the way you suggest this is not the reason to profile each image. The main reason for profiling each image is that the same piece of film scanned to the same resolution can have different grain structure because grain clumps are the result of not only the film structure, but also the varying colour and luminosity of what is being photographed, and how the film is exposed and developed. I only adopted a habit of custom profiling when I realized that one size does not fit all with negatives, so it pays quality dividends to custom profile and examine the results before moving on to the next stage.

As you say, we each do what works best for us, which is partly determined by the materials at play and what we are trying to achieve.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

dmerger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 680
NDQ - Scanning Resolution
« Reply #10 on: February 25, 2006, 03:49:30 pm »

Mark, I'll jump in here, even though your comment was addressed to Ray.  

If my understanding is correct, when you scan at about 3100 ppi with your Minolta, your actual scan is 5400 ppi which is then down-sampled to 3100 ppi.  So, with your method, you down-sample the initial "raw" scan.  With my method, I may down-sample, but after all my edits.  

I haven't done any tests to compare these two methods, but I expect that there would be very little, if any, difference visible in prints at 10x15 inches and smaller.  

I agree that there are disadvantages when working with very large files.   I often print larger than 10x15 inches, however, and the advantages of my method, for me, outweigh the disadvantages.
Logged
Dean Erger

dmerger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 680
NDQ - Scanning Resolution
« Reply #11 on: February 25, 2006, 03:53:06 pm »

Sorry, Mark, I was typing my last reply when you posted your last response.
Logged
Dean Erger

dmerger

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 680
NDQ - Scanning Resolution
« Reply #12 on: February 25, 2006, 04:54:21 pm »

Mark, I just ran a couple of tests with my Minolta 5400 (original version).  I ran the test twice and got identical times each test.  It took much longer to scan at 3060 ppi than it did at 5400 ppi.  That's not a typo, it took longer to scan at 3060 ppi.

I have very limited understanding of CCDs and how the Minolta hardware works, so maybe someone with more expertise can help.  I understand, however, that it's not physically possible for the Minolta CCD to have a continually variable ppi, just like a dslr can't physically have a continuously variable number of pixels evenly distributed across the sensor.  In other words, a CCD has a fixed number of capture sites.  Therefore, I can't understand how the Minolta scanner can physically scan at odd resolutions like 3100 ppi or 3060 ppi (and have those pixels evenly distributed).  

Mark, I guess that I wasn't clear in my post about my experience with Neat Image.  The profile I was referring to was a Neat Image noise profile, not a color management profile (which I've also made for my scanner).  As you point out, my experience is with Velvia and Provia only, not negatives.  Also, virtually all my film was processed by just a handful of professional labs, so I believe I have received consistently good processing.  Perhaps that consistency accounts for my success in using a Neat Image noise profile.  I don't know if I'd be able to reproduce that success with my negative film.  As you point out, there aren't any IT-8 targets for negative film, so using IT-8 targets wouldn't be an option in any event.
Logged
Dean Erger

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
NDQ - Scanning Resolution
« Reply #13 on: February 25, 2006, 06:54:26 pm »

dmerger, OK, I think I've explained why I don't use a standard noise profile.

Now, the scanner CCD business. I'm no technical expert about the innards of scanners either, so I can only go on what I've been told by people who are supposed to know. But what you say sounds sensible. The number of photosites is of course fixed. Question is whether the scanner uses all those photosites at any resolution. When I heard what I did about the effect of resolution settings, I thought that perhaps it has a hardware way of selective scanning off those sites, but on the surface of it, that would make less sense than software interpolation using all the sites.

Unfortunately, Minolta has shut down this business here and the guy I dealt with all the time is no longer with them, so I don't have a contact to revert back to and verify.

Anyhow, I think it doesn't matter. Whether the scanner does its thing by hardware or software or a combination thereof, I simply haven't found ANY upside to scanning more resolution than the maximum I think I shall need for any one image, because I would simply end-up rescaling or resampling with the above-mentioned issues. For reasons explained above I don't believe that the scanner resolution setting has a determinative impact on differentiating film grain except to increase the appearance of it the higher the resolution. Now for people who make much larger prints than I ever intend to, or use  lower resolution scanners, or some combination thereof, maxxing-out on the scanner resolution can make sense. For me it doesn't as a default option. There is also a fair amount of professional advice in the public domain suggesting that one should select scanner resolution according to the intended size of the output. But as a general principle everyone should do what they think works best for them regardless of what anyone says! Nothing like experimentation and making-up one's own mind based on what one sees out of the printer.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
NDQ - Scanning Resolution
« Reply #14 on: February 25, 2006, 09:24:06 pm »

Mark,
There is at least one piece of professional advice I recall (from Jack Flesher) which suggests there are sharpening advantages achievable by up-rezzing an image to a larger size than the intended print size before applying sharpening for that larger print size, and then down-rezzing to the intended print size using bicubic sharper.

You are no doubt familiar with the raging debates about whether or not any purpose is served in converting a native 8 bit image to 16 bit for processing purposes. The consensus appears to be that it is useful provided the image requires a lot of editing.

In general it seems to me that final print results could be better and not worse as you imply, if all editing is done with the maximum amount of picture information you can get, whether it's 16 bit as opposed to 8 bit or a 200mb file as opposed to a 40mb file. But I'd be interested in seeing some highly magnified crops demonstrating the differences that this expert body of opinion you refer to believe exist.

I now recall the circumstances of my own tests a few years ago. It was after I'd just bought the Nikon 8000ED to scan my MF negatives and slides. Someone on this forum stated quite categorically that better results would be achieved by scanning for the print size rather than downsampling a huge file in Photoshop.

Now a 6x9cm slide scanned at 4000 dpi in 16 bit (or 14 bit interpolated as the 8000ED appears to do) produces a really huge file in excess of 700MB.

I produced a number of A4 prints of the same slide, scanning with both methods and applying the same very basic processing to the image before printing. I printed all prints at 2880 dpi on Premium Gloss from 360ppi files on my dye-based Epson 1290. I really could not tell any difference between them. I resorted to lightly pencilling the scan method used on the back of each print and shuffled them like a pack of cards. Eventually, I did learn to pick up at least one small difference that would identify a particular print, without looking at the back. It might have been something like a minute and barely perceptible difference in the shape of the tip of one particular leaf in the top right hand corner of the print, or something like that.

I came to the conclusion that any differences that required such extreme pixel-peeping were entirely irrelevant. I was a bit annoyed at having wasted my time, ink and paper.  

As I've already mentioned, I can see good economic reasons for scanning only for print size and if you are in the business of of providing prints for clients from their film (and perhaps you are) then your method makes sense. If I could pay myself $40 an hour, I could make a lot of money rescanning my slides at different resolutions.  
« Last Edit: February 25, 2006, 09:46:10 pm by Ray »
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
NDQ - Scanning Resolution
« Reply #15 on: February 26, 2006, 12:03:03 am »

Yes, I remember hearing that idea about going up then down. I never tried it, because the methods I adopted using PK Sharpener Pro work just fine for me without the added steps, and as you say, if one really needs to pixel-peep to see the difference.....................

The 8 versus 16 bit debate is tangled-up with the colour space debate, and as you know ranges from abstract science to theology with occasional reference to tangible realities.

I think a consensus is emerging that if one works in very large colour spaces such ProPhoto, 16 bit depth is preferable because there is more data with which to fill the added space and this reduces the risk of banding and posterization.

However then the argument turns to the merits and risks of working in colour spaces as large as ProPhoto. The "narrower space people" do have a point to the extent one may not like how Photoshop handles gamut compression for certain colours that are way out of the printer's gamut. Sometimes, confronted with this problem, working in a narrower colour space happens to be quite effective at mitigating the worst effects of gamut compression on some colours.

But the flip side is the fact that printers like the Epson 4800 can produce certain hues that exceed the gamut of ARGB98. (It is not only the size but also the shape of the various colour spaces that matters.) There has been alot of discussion about choice of colour space in this Forum, and in others. I have more to show about it in the sequel to this article dealing with Silverfast, forthcoming. I frankly believe one needs to be pragmatic about these things and select whatever produces the most visibly satisfactory results, which will vary from image to image.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Anon E. Mouse

  • Full Member
  • ***
  • Offline Offline
  • Posts: 197
    • http://
NDQ - Scanning Resolution
« Reply #16 on: February 26, 2006, 02:32:03 am »

You will find digital cameras and scanners the same - they both work faster at maximum resolution because they do not have to process the image to reduce image size.
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
NDQ - Scanning Resolution
« Reply #17 on: February 26, 2006, 08:42:55 am »

AnonEMouse, image processing is so rapid in modern equipment that the differences of processing speed as a function of resolution should be small. One thing I never bothered to do is to clock it - on that matter in this Forum I was responding from memory, not scientific procedure, while dmerger has actually measured it and suggested my memory on that issue could be incorrect  -  we are both using the same scanner model. OK.

The underlying matter at issue here is whether different scanning resolutions do different things to the appearance of grain structures in the scanned image. They may well, but there are other factors that could be at least as if not more influential. The bottom line still emerging from my tests and results is that there are merits to selective resolution settings and merits to profiling the grain for each image. Having said that, anyone getting results in a more automated manner that are more acceptable to them, that is a win-win for them.
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."

Anon E. Mouse

  • Full Member
  • ***
  • Offline Offline
  • Posts: 197
    • http://
NDQ - Scanning Resolution
« Reply #18 on: February 26, 2006, 09:04:01 am »

Mark, I am not critisizing what you are saying. there just seemed to be a question of scanning time regarding resolution. I was just adding what I know about that point and nothing else.
Logged

Mark D Segal

  • Contributor
  • Sr. Member
  • *
  • Offline Offline
  • Posts: 12512
    • http://www.markdsegal.com
NDQ - Scanning Resolution
« Reply #19 on: February 26, 2006, 10:20:06 am »

Anon - no problem - actually I welcome criticism and debate - we learn new things that way.  

Cheers,

Mark
Logged
Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....."
Pages: [1] 2   Go Up