Poor scans - May explain why some switch to Digital

Landrum Kelly · March 25, 2009

Scott, there is an error on the link to the big file--or at least that is the message that I am getting.

--Lannie

rishij · March 25, 2009

Great, Lex, that's exactly how I figured it worked. Except I was assuming a magnification factor of 1 because I was assuming that a 4000dpi sensor actually had 4000 pixels per inch of the sensor... but, in fact, 4000dpi just refers how many dots per inch it's supposed to resolve on an image that spans the entire width of the CCD.

The Sony 5340 pixel CCD, for example, is only 21mm wide... That gives it an approx 4 micron pitch per pixel, and a magnification factor slightly less than 1 would be required.

Back to the Imacon...

Like Scott alluded to, you move the lens closer to the film for 35mm; this also dictates that the rear element of the lens be furthest away from the CCD. Move the lens away from the film and concomitantly closer to the CCD for larger formats.

One would think that for $20,000 Imacon coulda modified the mechanism so as to scan the MF in two slices, then stitch, as the Russian website Sergiy linked to alludes to. This way you could easily get 8000ppi scans from MF.

What I was thinking could actually give you a 2x resolution boost in both x & y directions, but would definitely slow scan times down. Still dunno if it's possible. I'm weary of mucking around with the CCD with my bare hands in a static-full environment :)

-Rishi

rishij · March 25, 2009

Yeah, Lannie, there's certainly quite a wealth of information in this thread :)

-Rishi

lex_harris · March 25, 2009

<a href="../photodb/user?user_id=940372">Scott Wilson</a> <a href="../member-status-icons"></a>, Mar 25, 2009; 02:11 a.m.

They are not real pixels and you are not seeing real resolution. I hope that people can get the idea that being able to read lines on a resolution chart does not come close to telling the whole story.

</blockquote>

Scott, perhaps we are misunderstanding one another. Earlier you implied that Mauro's assessment of resolution was not valid because line pairs were not discernible within a small window of pixels. What I am saying is that the assessment IS valid because the line pairs DO become discernible if you increase the window size. In other words there is in fact ordered data present, albeit very noisy data. Just because the data is noisy does not mean it isn't there. In your "noisy" image above, the faint line above the letters on the tyre remains visible over its full length, i.e. in the context of the overall image it is resolved and visible. According to your window test, it does not exist towards the outer diameter of the tyre. See below:

Here's the same crop with some NI applied. The line is faint, but visible:

Your 200MP image is certainly impressive and I agree there is more to an image than resolution alone, however I believe your "window" test is invalid as far as assessment of resolution is concerned and that was my point. Perhaps we have a different idea of what resolution means. In my view, if data is above the noise floor and discernible from adjacent data within the context of the image then it is resolved. Of course this is not the same thing as signal to noise ratio, which could be quite poor at the same time.

swilson · March 25, 2009

Lex, well to be more precise we are seeing images that optically resolve the detail, are not sampled high enough to fully resolve the image. In the case of film the sampling is random and so with a large enough area lines begin to appear. This tells us that the optics could resolve the lines, but unless you can still see the lines looking at a 1 pixel wide sample the image is not resolving down to the pixel level.

swilson · March 25, 2009

<a href="http://www.photo.net/photodb/user?user_id=423641">Landrum Kelly</a> <a href="http://www.photo.net/member-status-icons"><img title="Subscriber" src="http://static.photo.net/v3graphics/member-status-icons/sub8.gif" alt="" /><img title="Frequent poster" src="http://static.photo.net/v3graphics/member-status-icons/2rolls.gif" alt="" /></a>, Mar 25, 2009; 04:19 a.m.

Scott, there is an error on the link to the big file--or at least that is the message that I am getting.

</blockquote>

When you get past 100MP or so most browers can't handle it. The only way to view the image then is to download it to your computer and view it with software that can handle an image that large, I use eitehr photoshop elemets or ACDSee (ACDSee will sometime also have problems).

Most of the film scans that I see that are in the range of 200MP are so ugly at the pixel level that I would not want to print them at 300ppi. I figure that ideally a pixel should be a pixel and you should be able to take a 1800 x 1200 pixel crop out of it and produce a excellent looking 4x6 inch print.

Now just what one does with a 200MP image is another question, at print of that image at 300 ppi would be 35.7 x 62.3 inches.

Landrum Kelly · March 25, 2009

Scott, I finally got the 200MP file downloaded. (The error message did not come up this time at the start ofthe download.)

I am not only sitting here in awe right now. I also cannot but reflect upon the future of digital imaging--in or out of cameras of traditional size.

What is the future of film and film scanning? That question is unavoidable at this point.

In the meantime. . . .

--Lannie

swilson · March 25, 2009

For those who want to scan at very high resolutions

<a href="http://sales.hamamatsu.com/en/products/system-division/virtual-microscopy/products/nanozoomer-rs.php">http://sales.hamamatsu.com/en/products/system-division/virtual-microscopy/products/nanozoomer-rs.php</a><a href="http://sales.hamamatsu.com/en/products/system-division/virtual-microscopy/products/nanozoomer-ht.php"></a>

up to 110,000 ppi.

Scans up to 52x76mm

I wonder if they have run a piece of film through it just to see what comes out.

lex_harris · March 25, 2009

<a href="http://www.photo.net/photodb/user?user_id=940372">Scott Wilson</a> <a href="http://www.photo.net/member-status-icons"><img title="Frequent poster" src="http://static.photo.net/v3graphics/member-status-icons/1roll.gif" alt="" /></a>, Mar 25, 2009; 05:42 a.m.

Lex, well to be more precise we are seeing images that optically resolve the detail, are not sampled high enough to fully resolve the image. In the case of film the sampling is random and so with a large enough area lines begin to appear. This tells us that the optics could resolve the lines, but unless you can still see the lines looking at a 1 pixel wide sample the image is not resolving down to the pixel level.

</blockquote>

I guess we just have a different understanding of resolution. I'm happy to leave it at that.

Landrum Kelly · March 25, 2009

Scott, here is the "Danger" sign on the front end loader in your 200 MP crop.

Incredible!

--Lannie<div></div>

Landrum Kelly · March 25, 2009

And here is the "Warning" sign on the same 200 MP image of the front end loader.

Thanks for the hard work, Scott.

--Lannie<div></div>

sergiy_podolyak1 · March 25, 2009

As I understand they (photo industry) still do not understand exactly why and how photo grain is grouping under the light. In the future photo-science will find how and why, and then develop THAT SCANNER. It will recognize patterns in grain groups and open enveloped hidden info from grain. I suggest that knowledge will give photographers another 2X multiplication of resolution on film. So what? Shoot film, not digital. "Film Grain, Resolution and Fundamental Film Particles" 4.5 MB http://aic.stanford.edu/sg/emg/library/pdf/vitale/2007-04-vitale-filmgrain_resolution.pdf

Landrum Kelly · March 25, 2009

"As I understand they (photo industry) still do not understand exactly why and how photo grain is grouping under the light. In the future photo-science will find how and why, and then develop THAT SCANNER." --Sergiy Podolyak

</blockquote>

What will the market be, Sergiy?

--Lannie

sergiy_podolyak1 · March 25, 2009

Do you, Lannie, mean THE SCANNER will be too expensive or too big? I dont think so. If computers passed their way from 1000 tubes/10kW/1000 operations per second/nuclear bomb calculations to 1 billion transistors/1 sq.inch/1W/1 billion operations per second/everywhere, so I dont see reasons why this path is not suitable for "photography"? There could be very unexpectable applications of photo-art in the future, and THE SCANNER ... will ALSO scan old 35 mm films.

Landrum Kelly · March 25, 2009

I asked a question, Sergiy. I made no assertions.

I expect scanning technology to continue to grow and improve, but whether it will proceed by the path you suggested is not something that I can answer.

--Lannie

sergiy_podolyak1 · March 25, 2009

Lannie, I have heard a bit of sarcasm in your question, so I thought have to assure you in a bright future of photo-art. You wrote : "...is not something that I can answer". No problem, the other 50% of the answer of that question is well known - electronic digital cameras ARE STUCK (stopped) with modern resolution*dynamic range=const. And there will be no progress any further. There are not a one but many reasons for that. Electronic engineers know that very well. But sadly they do not explain that to public. A very interesting posting from another thread of this forum, need to be posted here: http://www.photo.net/film-and-processing-forum/00ROOo ........................................... Ken Kryda , Nov 24, 2008; 05:22 a.m. "I've been designing digital cameras for about 30 years now, including the first 2 Leaf camera backs, and film scanners for Associated Press and most recently, on the opposite end of the quality spectrum, camera phone sensors and image processors at Micron. I started out as an amateur photographer in the late 70's with film but worked on digital systems as a career. Today I shoot medium format B/W with my Hasselblad 503 and develop my own film for fun and use a Canon 5D for 360? HDR. At Micron I served on the I3a (International Imaging Industry Association) CPIQ image quality standards group. So most of what I can contribute to this discussion is more on the engineering side, though obviously I also use film for some very good reasons. The issue of image quality and the differences between film and digital is something I'm deeply immersed in so, sorry if I'm too verbose. RESOLUTION To start, I've measured the resolution of many systems with a few different methods and would say that in the 35mm format, digital is just as good if not better than film. However there are more dimensions to this issue than just pixel count or grain. The pixel size and the light level affect resolution a great deal. As pixels get smaller they have less surface area to collect light and produce more noise. To compensate for this, noise reduction algorithms are used but often blur the image in the process. DSLR's usually stay above 5 um (micron) pixels but still use NR which can blur or add artifacts. Midrange DSCs can go down to the 2um and < 3MP camera phones are less than 2 um. The demosaic algorithm that converts the RGGB Bayer pattern to an RGB picture element. Note, a "sensor element" is not a picture element but both are called a pixel and there is a 300% difference. (http://dictionary.reference.com/browse/pixel ) The demosaic can have a 15% variation in resolution, with or without color artifacts. I don't call this color interpolation as many do because interpolation (bilinear) is the worst form of demosaic. Most use pattern recognition to help "synthesize" pixels. The best I've seen is the Hirakawa Parks algorithm which yields no color aliasing and very high resolution with no "optical low pass filter". I developed a metric for resolution for the I3A international standards group that is the optical percentage of " the-number-on-the-box" at MTF50 (half contrast). It should be deployed in the industry next year. On average, a digital camera only gives you about 33% of the stated number of pixels. A really bad camera can go as low as 20%. The crazy thing about this is that a camera with more smaller pixels can actually have less resolution that a camera with fewer larger pixels. And the resolution gets lower and lower as the light decreases. This is why I prefer a camera with larger pixels even if it has less resolution. The Canon 5D has 8um pixels and I would prefer it over a 16M with 5um pixels. If I set the camera to ISO 1000, capture raw, turn off the NR in the raw software (PhotoShop) and use a transform based noise reduction algorithm like Noise Ninja, I can do night shooting at 1/125 sec with my f1.2 and have high resolution with now noise and no blur. This can not be done with film. Also note that sharpness and resolution are 2 very different things. Our brain reacts most to edge sharpness which is easily 1/10 of the resolution, or the maximum frequency (detail) the camera can capture. In digital, sharpness can be "turned up" and actually look better than higher resolution images with less sharpness. When I became interested in the greater resolution of medium format film, I asked a few professional scanning services how many "mega pixels" I could get from scanning medium format. No one could give me an answer, no one. They'd tell me how many dpi they offered but, having designed film scanners, I new that was not the total optical through put of the system. So, I jumped in, bought a used 503, took a picture of an ISO12233 resolution chart at 1/5 frame, had it developed at a professional photo store and looked at it under a microscope. As I looked at smaller an smaller lines there was a amazing amount of "dirt" and could only see the equivalent of 30M pixels (line size = dirt/grain). I assumed that film was so uncommon that their chemicals we not fresh. So, I developed my own film with Kodak D70 and got about 40M pixel. I then moved to XTOL (about $1 more) and got around 50M pixel. I though, wow an extra 10M pixel for $1, what a deal! I knew that for about $60K I could get a digital back that would give me the same resolution but, I didn't have $60K. I guess I've spent about $8K on my Hasselblad body and lenses and can get about 50M pixel. A Canon DS MKII was also about $8K then and only 16M pixel. DYNAMIC RANGE One of the big advantages of film has been the wide dynamic range of light it can capture but some of the better DSLRs can do just as well if you're careful. Average DSCs get around 9 stops. And yes, pixel size makes a big difference. Here's some comparisons: http://www.clarkvision.com/imagedetail/dynamicrange2/ You can see that film seems to have a larger range but the noise (grain) becomes a factor and can not be processed in the same way as digital. I shoot Fuji Acros (ISO 100) in my 503 and I think I get about 13 stops but I haven't measured it yet. As we know from Ansel Adams, 10 stops is OK for most outdoor scenes. COLOR Many people have said they like the color of film better than digital, but those people also admit they haven't spend a lot of time working with color processing software. I would suggest looking at DxO's "film looks": http://dxo.com/intl/photo/filmpack/available_film_looks In theory both are tri-stimulus color systems and as long as you have enough dynamic range in the color channels (>12 stops), you should be able to create what ever color look you want in digital. I've also developed a high dynamic range, hyper-spectral printing process using an HP 3100, rip engine and 2 layer transparencys with a xenon arc light source( 30cd/m2). The color was measured with a Spectroradiometer and found to be the exact same as the real scene in nature. The images were captured with a Canon DS MkII (14 exposures). So, I'm pretty confident you can get the same color quality with digital but it may be more work than just pressing a button. " ------------------------------------------------------------------------------------------------------------- He said only 30% of the truth about electronic cameras. The rest 70% is even worse.

rishij · March 25, 2009

Film scanning technology will not grow & improve because there's no money in it , and companies do not prefer business models with diminishing returns.

Look at the economy now. Because of the crash, even certain 'higher end' tech products with smaller markets are all 'backordered' or 'not available until an unknown future date'.

They've already got drum scanners and movie reel scanners that hold film flat & provide pretty good resolution with PMTs. But even the movie industry is shifting to digital... save for a few directors brave enough to film in 70mm.

The push for higher end affordable desktop scanners, though, is just not there. Because the small market is not incentive enough for companies. Look at the film scanners bein released these days-- they're 'pocket' pieces of crap that scan 1-2MP resolution scans from your negatives. That should be an indication that companies are only willing to produce what will sell... and to the mass public, unaware of the inherent resolution of film, a $100 device that'll extract their family pictures/memories is good enough for them. Not a $1k or $2k or $4k high-resolution scanner that you have to spend a couple years learning how to use anyway to get the optimal result.

What might happen, though, is some individual(s) who, with enough time, resources & money, end up building such a device. For example, the digital camera setup that Sergiy linked to. With enough modifications, one could actually build a viable device out of that.

-Rishi

Landrum Kelly · March 25, 2009

"Lannie, I have heard a bit of sarcasm in your question. . . ." --Sergiy

</blockquote>

My question was "What will the market be?" If you read sarcasm in that question, that is something that you read into it. I was not being sarcastic, although, yes, I was challenging you.

What will the market be? Will it be big enough to justify the expense of the research, etc.?

I do not know.

--Lannie

rishij · March 25, 2009

Ahem, does the slide digitizer employing a dSLR & extension tubes + macro lens not look distrubingly similar to one of the scanners we've been discussing in this thread... (see below)?

Like I've already said, at $20k, the Imacon's a rip off. If it weren't for my Ph.D, I'd be designing something like/based-off the digitizer on the right...

-Rishi

mauro_franic · March 25, 2009

Rishi, I used to have a target resolution slide but I can't find it. I will look for it again tonight.

I think I bought it here.

http://www.aig-imaging.com/mm5/merchant.mvc?Screen=SFNT&Store_Code=AIIPI

robert lee · March 25, 2009

Actually, all you'd really need is to add a digitally controllable X-Y table. Shoot tethered and do the rest automatically under software control. A CS mount type USB industrial inspection camera can be easier to work with.

Take multiple exposures to remove sensor noise, increase digitization dynamic range. Stack exposures taken at different focus points to remove film flatness issues. Flat stitch the small sections together for an arbitraily high equivalent digitization resolution - want a 10k dpi film scan, digiscope a microscope.

By the way, use an old slide projector for a really good spectrally consistent, ultrabright light source. Bounce it up with a mirror and diffuse with transluscent perspex.

mauro_franic · March 25, 2009

Nice dreaming.

The Coolscan 9000 is still an awesome capture from medium format with a little effort an affordable price.

In the rare event one needs more information of a particular 6x7 frame than the coolscan 9000 can capture (hard to think of such application), it can conviniently be sent out for drum scanning.

rishij · March 25, 2009

Ah, I see Mauro. A target slide printed like that would have a much higher contrast ratio than the Velvia you sent me, where there's already been a lot of contrast lost due to print + lens + film. Hence your estimation with the target slide would be much higher.

As would mine, for the Nikon & the Minolta & Imacon of course! If it's not too expensive, I'll get myself one.

-Rishi

rishij · March 25, 2009

Robert Lee,

Couple of thoughts: digiscoping a microscope wouldn't lead to good scans... look at Mauro's microscope capture. The light source is all wrong and, with the condenser, results in very harsh directional lighting that accentuates bubbles & dust. Popping a diffuser in the light path only helps so much.

Re: the light source, I found CCFL much better than point light sources (halogen, LED) that have been collimated using, for example, two plano-convex lenses... just because the light source is more evenly distributed to begin with and, therefore, more amenable to diffusion using some diffuser in the light path. But I'll have to take apart a slide projector & see the condenser setup... maybe it'd work well.

There are high-precision X-Y controllable tables... yes that's exactly what I was getting at. One could use these to achieve higher resolution scans without stitching... think about it.

And Mauro -- make up your mind! Do you want a higher resolution MF scanner or not? :)

-Rishi

swilson · March 25, 2009

My dream scanner would, handle MF, scan very fast, and have ICE and cost less then $500.

When I say fast I am thinking in the range of 4-5 seconds for a 35mm frame and 10-20 for

MF, with ICE on.

If there was a large market for such a scanner I think it could in fact be made for less then $500.

My feeling is that the market for film scanners has never been large and is shrinking pretty fast, so I don’t expect anything really new in film scanners.

Sign In

Poor scans - May explain why some switch to Digital

Recommended Posts

Link to comment

Share on other sites

Top Posters In This Topic

Popular Days

Top Posters In This Topic

Popular Days

Posted Images

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Link to comment

Share on other sites

Create an account or sign in to comment

Create an account

Sign in