Reducing granularity in 35 mm film scans

<p>Hello, all, from a new member.<br>

I hope someone can offer guidance with a really vexing problem. It shows up mainly in continuous tone areas like skies, and does not seem to be limited either to B/W, color neg, or slides.<br>

My practice has been to scan at 4,000 ppi in my Nikon Coolscan V ED film scanner.<br>

If I understand correctly, what I am seeing is a sort of digital "artifact" resulting from the scanning process called "grain aliasing." <br>

It is truly ugly, as it bears no resemblance to film grain, and even the finest grain B/W films are affected by this in the scanning. Worse than ugly, it seems to defy any noise reduction programs I've tried. What concerns me most is that even images shot on the finest grain films show this ugly pattern on my calibrated monitor.<br>

Recently, I've started using the Enhanced Scans done by North Coast Photographic Services, thus getting the same large 4,000 ppi scans as I had been doing one by one on my Nikon scanner. They are sharp and the colors are lovely, as well as producing excellent B/W tonality. But the granularity pattern is there, just as in my Nikon scans.<br>

When I subsequently edit, either in Corel Paint Shop Pro Photo or PS Elements, I first change the resolution of the large TIFF (or in the case of NCPS, jpg) file from 4,000 to 300 ppi, while keeping the scanned pixel dimensions of ~5,500 x ~3,300.<br>

This does not in any way change the granularity patterns on my monitor. Curiously, when I print such images on premium photo paper (full-frame on 8.5 x 11 sheets), the granularity is not as noticeable, but I remain concerned as my edited images are sometimes professionally printed to much larger dimensions.<br>

Can anyone offer some layman guidance as to what is going on here - AND recommend an easy, straightforward stand-alone program that reliably removes such granularity from continuous tone areas.<br>

I've been behind one lens or another for decades, but consider myself somewhat a newbie to scanning. Nonetheless, I've lately made a major re-commitment to shooting film, so between new work and thousands of legacy slides and negs, I am definitely looking for a way to remove this particular headache from my film-to-digital workflow.<br>

Many thanks!<br>

Steve</p>

<p>If your problem is indeed grain aliasing, it is like any other type of digital aliasing -- specifically, high frequencies (... ie, spatial, audio, whatever...) get wrapped around the sampling frequency and appear as spurious low spatial / audio frequencies. </p>

<p>Once this happens, no amount or type of post-digitization filtering will cure the problem without reducing everything to a big blur / acoustic mush. The problem *must be* attacked before digitization. Specifically, you must reduce the extremely high frequency, out-of-band spatial frequency content of the original before it is scanned. Usually, this can most easily be done by very slight defocusing of the optics of the scanner. Obviously, one does not want to defocus so much that it will be seen as a reduction of the resolution of the scan, so we're talking about very fine adjustments of the optics.</p>

<p>HTH,</p>

<p>Tom M</p>

<p>Hello Tom,<br>

Thanks for your response. As an audio professional, I do indeed understand your analogy to digital artifacts relating to a sampling frequency. Nasty stuff when listening to real music...<br>

So, I guess I must - first - determine if what I'm seeing in my skies REALLY IS grain aliasing. Perhaps it is some other by-product of the digital scanning process. Otherwise, I have to think more film photographers would be pitching the same gripe.<br>

I sure hope it's something else, as I haven't a clue as to how to slightly de-focus my Nikon scanner...nor would I imagine would NCPS care to de-focus their professional Noritsu Koki QSS scanners.<br>

So, with thanks and respect to Tom, has anyone else encountered this in their scans from 35mm film.<br>

Hopefully,<br>

Steve</p>

<p>Yes, but in my case it seemed more attributable to the film being slightly overexposed and/or the developers used.</p>

<p>Steve, you should post an example.</p>

<p>Hi Steve - Some scanners have software control of focus ... even accessible through scanning programs like Vuescan.</p>

<p>Lacking that, one could always set the film holder a bit off of the optimal height, or intentionally place a piece of suitable, mm-thick, anti-reflective glass under your negative. I also have heard of a thin transparent plastic with an orange peel textured surface (ie, much like a some soft focus filters) that can be placed under your negative to kill the perfect focus and presence of very high spatial frequencies. Unfortunately, i don't remember the name or mfgr of that product.</p>

<p>Cheers,</p>

<p>Tom M</p>

<p>Btw, what Tom is saying just here is comparable to an anti-alias filter. If the problem is, as we assume, HF/digital bandwidth interference, it should do the trick.</p>

<p>On the other hand, the fact that few people are confronted with this problem casts a doubt on (y)our diagnostic.</p>

<p>Also, grain is (or at least should be - what film do you use ?) random in nature. It is comparable to audio noise, in that it has no frequency to interfere with digital resolution. Noise detail that is finer than the sampling capabilities should statistically cancel out, shouldn't it ?</p>

<p>I am also a sound tech, or at least that was my education. That time is long past now, so maybe I don't remember the theory so well... Tell me, Stephen, am I making sense here ?</p>

<p>"my edited images are sometimes professionally printed to much larger dimensions."</p>

<p>I'd suggest actually comparing* larger prints from your scans, the professional scans, and wet projection printed 35mm at similar dimensions. My guess is that what you see will be very similar. My darkroom experience was a long time ago, but I was never happy with 11x14 prints from Panatomic X, whereas Plus X in 120 was gorgeous at that size. More recently, what came out of my Nikon 9000 looked a lot like what I saw in the film with a 30x microscope. When grain aliasing was first discussed, it looked to me to only be a problem with certain (but widely used) consumer films. So my opinion here is that what you are seeing actually is the grain clumping. Still, scans of even ISO 100 color negative films could be pretty gritty, so I just stuck to Provia 100F.</p>

<p>*: This is because looking at a 4000 ppi scan on screen is a very detailed view of the film: it's more than a 20x loupe, probably less than a 40x microscope.</p>

 

<p>Thank you all for the attempts to help. And Dan and Les, you are so right - it would have made more sense to post a photo showing what I'm yakking about! So please see the large-image link below for a good illustration - not saying wonderful photo - just good illustration ;-)<br>

Les, to answer your question, I've seen this crude sort of granularity on Ilford XP2 Super (ISO 400), but I've also seen it on Agfa Pan 25 which is about as fine grain a film as I know of. As I say, it doesn't seem to relate so much to native film grain as to some sort of digital over-sharpening in the scanner although, I hasten to add, I am not using any enhancement modes in the scanning process.<br>

Dang, this is frustrating.<br /> Thanks again, gents, for your willingness to help me sort this out. Photography is supposed to be fun...or so I thought before scanning entered my workflow.<br>

Steve</p><div></div>

<p>Looks pretty normal for 35mm. You haven't indicated the type of film , though.<br>

I have had good success by making a new layer, using gaussian blur, and then recombining the original image and sharpening. This can be done selectively.</p>

<p>This is why I shoot medium format. With tmax 400, you don't see any grain.</p>

<p>FS: <em>"On the other hand, the fact that few people are confronted with this problem casts a doubt on (y)our diagnostic...."</em></p>

<p>Actually, I don't think that is a true observation. Several years ago when lots of folks were busily digitizing their archives there were tons of reports and of grain aliasing. Google the term (with quotes around it) and you will almost certainly come up with large numbers of posts and websites discussing this topic. A classic is: http://www.photoscientia.co.uk/Grain.htm</p>

<p>Another effect that contributes to undue enhancement of grain is the Callier Effect (Google it) </p>

<p>FS: <em>"...Also, grain is (or at least should be - what film do you use ?) random in nature. It is comparable to audio noise, in that it has no frequency to interfere with digital resolution. Noise detail that is finer than the sampling capabilities should statistically cancel out, shouldn't it ?..."</em></p>

<p>Unfortunately, the answer is "no, it won't cancel out". Suppose, in an audio system, the Nyquist frequency is 10 kHz (never mind the lack of realism with this number - I just wanted a round number for the example which follows), and you have random noise uniformly distributed from 10 kHz to 20 kHz. Noise around 11 kHz will come out at 9 kHz, noise at 12 kHz will come out at 8 kHz, etc. etc. Thus the out of band noise contributes mightily to the power spectral density under 10 kHz. In other words, it make the system seem to have a higher, program dependent, wide band noise floor. </p>

<p>HTH,</p>

<p>Tom M</p>

<p>A diffusion filter can help. http://www.allari-photo.com/scanhancer.html<br>

http://www.scanhancer.com/ can make a custom size for your Nikon.<br>

A quality slice of frosted glass will do much the same...</p>

 

<p>Tom: Thanks for the refresher :-) !<br>

Do you think that's what happened to Steve's picture here, now that he showed it to us ?</p>

<p>Steve: btw, reducing any resolution/size parameter other than the actual pixel dimensions won't change anything. (except, of course, that picture flaws are usually less noticeable on smaller prints ;-))</p>

<p>A mild case of reticulation maybe? Do you have minimal control of the post-devloper bath temperatures?</p>

<p>Greg - Your suggestion for the diffusion filter is good and will definitely help (on the appropriate scanner - see the mfgr's website for more details). If I understand it correctly, it goes on the light source side of the film, not between the film and the optics / sensor. If so, it will help with the Callier effect, but not with grain aliasing. What I was referring to was orange-peel / pebbly surfaced glass (aka, anti-Newton ring glass) that you could put on the sensor side to roll off high spatial frequencies if the focus can't be manually adjusted.</p>

<p>Fred - re your question about what I think is going on in Steve's image, I agree completely with Les' comment on over sharpening. In addition, it looks like the contrast was pushed too much, early in the processing chain. Whether or not this was done by chemical means or in software, or both, I can't say, but typically, one never wants to overcook the sharpening and contrast early in processing. </p>

<p>I think there is a virtual certainty that the Callier effect was involved, but whether or not grain aliasing is involved is anyone's guess until the aspects I just mentioned are fixed. The only way to determine if grain aliasing was a major component in the problem would be for the OP to do another scan, processed absolutely identically to the one posted, except set the scanner focus slightly off and look for a disproportionate reduction in grain. </p>

<p>Cheers,</p>

<p>Tom M</p>

<p>Konrad, your article is a summary of well-known, post-digitization NR software. Sure, such approaches will smooth out an image, but (a) if the problem is grain aliasing, the only real solution is to deal with it before digitization. This is fundamental to all anti-aliasing approaches whether you are dealing with audio signals on your cell phone, the design of a modern digital EEG machine, or an electro-optical imaging system for a military customer. The necessity of filtering out high frequencies before digitization is so fundamental to modern electronics that it is typically taught in junior year undergraduate electrical engineering courses. Current A/D converters regularly include extremely high order (ie, sharp-cut) low pass anti-aliasing filters and include features like oversampling followed by decimation in order to reduce this problem in time-series data.</p>

<p>In addition, from the previous discussion in this thread, the OP and the other respondents sound like they are quite knowledgeable about conventional image editing software, and almost certainly know about both built-in and plug-in NR tools. I'm quite sure they are looking for something a "bit more" on this topic, although for a novice reading this thread, the recommendations in your article could be useful.</p>

<p>Tom M</p>

<p>Sincere thanks to all who addressed my issue. It's truly illuminating to see the kind and spectrum of intelligent responses on this forum. Must say, some of the advice was well above my head in terms of technical theory - and perhaps I'm to blame for that if my use of the term "grain aliasing" was an inaccurate description of what I was seeing in those scans.<br>

Thanks to all who looked at my scary-grain example - I'm going to HOPE that Les was correct in his comment "Your example scan is aggressively over sharpened by default on a Noritsu machine as seen in the clumpy look of the grain. This is certainly hard to post process." Amen to that!<br>

And thanks to Konrad for the tutorial, which I absolutely intend to follow.<br>

Guess it comes down to when your left headlight is out, the first thing to check is the bulb and not the engineering blueprints for the car's electrical system or the theory of computer control ;-)<br>

And Les, thank you for posting your useful scans - including the Tri-X image. Also for the absolutely rational advice to scan the same negs I got back from NCPS and compare the results between their possibly aggressive Noritsu and my lowly little Nikon Cool Scan V ED. I'm guessing that will prove instructive. Perhaps it was too much to expect that outside scans from entire rolls would prove as workflow-easing as just scanning the shots I'm really interested in at home.<br>

We shall see. Meantime, I have a project that's eating up my spare time, plus an upcoming trip - during which I shall do the FUN part of photography and shoot more film over which to obsess and swear when it's time to post or print ;-)<br>

When I get back, I intend to follow the "basic" advice described above in hopes there will be a "scanning epiphany," and will report back on what shakes out - hopefully that ugly clumpy looking grain you were all so kind to address.<br>

Thanks again - nice to be connected with such a pool of experts.<br>

Steve</p>

<blockquote>

<p>...if the problem is grain aliasing, the only real solution is to deal with it before digitization.<br /><br /></p>

</blockquote>

<p>But is it grain aliasing? Tom, if my memory is correct, you have also worked with camera scanning (to use <a href="http://www.dpbestflow.org/camera/camera-scanning">Peter Krough's terminology</a>). Have you compared scanning the same transparency with a scanner against camera scanning? For grain reduction? With camera scanning you can do a lot more with diffusing. Peter Krough, for example uses three layers of diffusion: A large softbox with two layers of diffusion and a translucent white sheet of Plexiglass in between the softbox and the film carrier. I don't think you can get the same amount of diffusion by slipping a single layer of frosted glass in a conventional scanner.</p>

<p>Peter states that camera scanning is better across the board than any kind of conventional scanner and implies that it is time to retire conventional scanners. I think that grain deduction via diffusion is one of the reasons why (Resolution is another reason and he put up <a href="http://thedambook.com/scans/">high resolution TIFFs to demonstrate</a>.)</p>

<p>I've been experimenting with camera scanning, but I am in the process of assembling all the pieces, so I don't have any samples to show myself.</p>

<blockquote>

<p>Peter Krough...</p>

</blockquote>

<p>"Krogh", not "Krough." I thought I had corrected this, but apparently got confused with the meanings of the "Confirm" and "Update" buttons. Sorry.</p>

 

<blockquote>

<p>You only have to download the files you reference to see that a Canon 1Ds MKIII (<strong>$7,500</strong>) does not do as well as a 4000dpi Coolscan 8000 (<strong>$2,000</strong>) scan of the same image.</p>

</blockquote>

<p>The article was written a few years ago. The 1DS MKIII is 21 MP and only has <a href="http://www.sensorgen.info/CanonEOS-1D_MkIII.html">11.5 stops of dynamic range</a> A better comparison would be against a <a href="http://www.bhphotovideo.com/c/product/842926-REG/Nikon_D800_D_800_SLR_Digital_Camera.html">$3,000 36 megapixel Nikon D800</a> that has <a href="http://www.sensorgen.info/CanonEOS-1D_MkIII.html">14 stops of dynamic range</a>. And can be used for more than scanning.</p>

 

<blockquote>

<p>BTW the overprocessing conducted with the DSLR shows JPEG artifacting very similar to the Steve's originally posted Noritsu issue.</p>

</blockquote>

<p>The 1DS file <a href="http://thedambook.com/scans/">he supplies</a> is a DNG file so there shouldn't be any JPEG artifacting. Did I overlook the Noritsu issue you are referring to, or is this in a different thread?</p>

 

<blockquote>

<ul>

<li>Of course the Canon 1Ds MKIII's full resolution is maximized on the 35mm sample while the Coolscan 8000 applies the 4000dpi up to MF - doubling the resolved detail even more.</li>

</ul>

</blockquote>

<p>It is true that MF film has more resolution than 35mm film. But I would like to see a comparison against a Nikon D800.</p>

 

<blockquote>

<ul>

<li>The Canon 1Ds MKIII will require multiple captures to maximize latitude.</li>

</ul>

</blockquote>

<p>But it only has 11.5 stops of dynamic range. Even film is hard pressed to reach the 14 stops that a D800 can achieve.</p>

 

<blockquote>

<ul>

<li>Scanning color negatives with the Canon 1Ds MKIII is not optimal for color processing.</li>

</ul>

</blockquote>

<p>There is no intrinsic reason why scanners can correct negatives any better than camera RAW files can be corrected. Bits are bits. I have been scanning negatives with both my 60D and with my Plustek 7600i, using VueScan to process from each, and find that I have to struggle as much with my 7600i scans as I do with my 60D scans. Yeah, a Plustek 7600i isn't a Nikon scanner, but a comparison on Luminous Landscape showed <a href="http://www.luminous-landscape.com/reviews/scanners/plustek.shtml">that it isn't far behind</a>. (Optically.)</p>

 

<blockquote>

<ul>

<li>Coolscan ICE is very useful and not available with DSLR scans.</li>

</ul>

</blockquote>

<p>It is true that ICE can be useful, but if your negatives aren't all scratched up or covered with mold and fungus, then a puff of air removes pretty much all the dust, when doing camera scanning with a properly diffused light source. Scanners are horrible at exaggerating grain, dust and scratches. Which is where this thread started.</p>

<p>I would like to see more tests of camera scanning, preferably done with more modern equipment. I have a 60D, which has less than 11 stops of dynamic range, and is getting a bit long in the tooth. I am still working on assembling good film carriers. From my experiments with sub-optimal film carriers, my Plustek 7600i looks like it has about the same resolution as my 60D scanning at 3600 PPI and a bit more resolution scanning at 7600 PPI. And I know that Nikon scanners have a bit more resolution than my 7600i does.</p>

<p>But my 60D does way better at reducing grain, dust and scratches. Because of diffused light, so there is less need for ICE type noise reduction. But again, I am still working on optimizing my film carriers, so I don't have a definitive conclusion myself--my 7600i excels in some areas and my 60D excels in others. Which is why I wanted to hear from anybody else that has tried camera scanning. Because they don't make Nikon scanners anymore, while DSLRs are improving by leaps and bounds. (I can't justify a D800 (or D800E) myself as a camera, but if somebody demonstrated that it produced superior camera scans, then it would be a lot more compelling to me.)</p>

<p>Thanks for sharing your test results. I can see that your combination of camera + lens has less resolution than your scanner. But there are other combinations of cameras and lenses...</p>

<p>First of all, I owe an additional apology. I had made numerous edits to my original post before posting, in addition to correcting the misspelling of Peter Krogh's name. My post as posted contained a paragraph that started with "Peter states that camera scanning is better across the board than any kind of conventional scanner" After writing this (but before committing the post), I went back and reread the Krogh "camera scanning" pages and realized that he stated no such thing. Rather, his point was more along the lines that camera scanning can provide results that are good enough for some applications, and that camera scanning is much faster than using a conventional scanner. I rewrote my paragraph to reflect this and then pressed "Confirm" (without first pressing" "Update"), with the results being that all my edits were discarded. Again, I apologize for unintentionally misstating what Peter Krogh actually wrote.</p>

<p>My own primary focus is digitizing my family's negatives and I am not looking for maximum resolution, so an image that makes a sharp 5x7 (or half-decent 8x10) is entirely acceptable. For my purposes. Based on my experiments to date, I am reasonably confident that camera scanning with my 60D will meet my needs. Right now my sticking point is that I use a PhotoSolve Extend-a-Slide and I have problems with the Extend-a-Slide 35mm negative carrier, so I am in the process of fabricating my own. But I have processed a boatload of 35mm slides with the Extend-a-Slide carrier and am happy with the results. And processed 20 or 30 Instamatic 110 negatives with good results (a lot better than the original prints.)</p>

<p>I never shot film myself. I recently bought several film cameras on eBay and am in the process of shooting my own negatives under controlled circumstances. (Use a tripod, shooting the same scenes with both my 60D + 17-55 f/2.8 lens and two Canon Elans (to eliminate the effects of a single camera having AF problems) with a Canon 40mm pancake lens. My first round of test was spoiled because I took the film to my local Walgreens to be processed and they had bad chemicals. After I pointed this out to one of the photo lab technicians, she agreed that there was a problem and to not bring film in until they have the film developing chemicals replaced. Once this has been done I will shoot another round of tests so I can cross compare everything. In particular, comparing the camera scanned negatives against how the 60D shot the same scene directly. And also against how my Plustek 7600i scanner scans the negatives. (The 7600i is (optically) almost as good as a Nikon scanner, according to the Luminous Landscape comparison review I cited in my earlier post.)</p>

 

<blockquote>

<p>You do of course understand that the higher resolving the "scanner" gets, the more obvious dirt and scratches will be - no getting around it.</p>

</blockquote>

<p>It is also been known for about many years that undiffused light amplifies the effect of grain, dust and scratches. Ever since photographers started using enlargers. Which is why a lot of enlargers let you add a diffuser. (Or use undiffused light to partially rescue a slightly out of focus negative.) I am still indeterminate about the effects of additional diffusion, because I still need to fabricate transparency carriers for my Xtend-a-Slide that will let me experiment with differing amounts of diffusion. And do the tests I described, above.</p>

<p>I realize that I have adopted several points of view. For my current purposes of digitizing my family's photo archives, I am happy with the sharp 5x7 level of quality. But I am also interested in what can be achieved with SOTA everything camera scanning. Or with the same level of camera equipment as I currently have, butwith a different choice of bodies and lenses. (The 14 stop Nikon D7000 has my attention.)</p>

<p>Now I <em>think</em> I have "Updated" all my edits to this post so they are frozen. Now I'll press "Confirm" and see if they really are.<br>

Wayne</p>

 

<p>Well, from my perspective, "camera scanning" is the method of choice for me when copying my slides and b&W negs. Much less granularity and dust and pretty well equal resolution and much quicker, and with Lightroom, easier color adjustments. It's no contest. Proviso is you need a full frame system with a 1:1 optimized macro lens and, in my case, a dedicated slide copying set up including diffusion tungsten light source. Dust spotting is still necessary, but no big deal for most shots. I only use my Canon 4000FS for color negs now as Vuescan does the initial color adjustments which are difficult in Photoshop.</p>

<blockquote>

<p>I only use my Canon 4000FS for color negs now as Vuescan does the initial color adjustments which are difficult in Photoshop.</p>

</blockquote>

<p>VueScan can open JPEG, TIFF or RAW files so you can process negatives that were camera scanned just the same as if they were scanned with a conventional scanner. With one caveat: you might need to use something like a <a href="http://www.bhphotovideo.com/c/product/44119-REG/Rosco_RS320211_3202_Full_Blue.html">Rosco # 3202 Full Blue tungsten-to-daylight Color Conversion Gel Filter</a> to remove the orange mask. I have better luck with VueScan color conversions when I shoot through a #3202 filter than I do if I try to convert a straight shot (with uncorrected orange mask) of a negative. (I started testing from a <a href="http://www.bhphotovideo.com/c/product/45184-REG/Rosco_950SBCNG0103_Cinegel_Swatchbook.html">Rosco Cinegel Swatchbook</a> and the #3202 was the first one in the swatchbook and seemed to work, so I ordered a full sheet from B&H.)</p>

<p>Others have reported good luck converting negatives with <a href="http://asf.com/products/plugins/rocpro/pluginROCPRO/">Kodak ROC</a>. But I already had VueScan Pro and it works as well with camera scans as it does when scanning negatives with my 7600i.</p>

<p>I converted over one hundred frames of Instamatic 110 negatives that were camera scanned this way, using VueScan to open my Canon 60D RAW files, using a partially home made 110 negative carrier (because my Plustek 7600i can't work with 110 negatives.)</p>

<p>I've posted about this in several threads so I won't say more, other than I am in the middle of doing tests comparing scanning negs with my camera scanning setup with scanning with my 7600i. One thing I discovered last night is (surprise) negatives are negative. Meaning that camera noise that normally shows in shadows, shows up in highlights in converted negatives. Which translates to noisy skies unless I am careful with keeping ISO low and nailing the exposure. Thinking about it, this should have been obvious, but I didn't notice it until I did my first A/B comparison of scanning the same negative with my 60D and with my 7600i.</p>

 

<p>Wayne,</p>

<p>That's good to know about color negative scanning - I had not thought of that. Actually my Chromapro has all the necessary dichroic filters to produce (in theory) a direct positive when scanning. I haven't tried it yet though. I'll probably try it with Vuescan as you suggest. I think you are correct that a gel filter for tungsten to daylight may be a good idea. Good tip! "Scanning" black and white negs with the copying system certainly works great.</p>