Poor scans - May explain why some switch to Digital

[Landrum Kelly]

<p>Rishi, even the color tones are better! Holy cow! What one can learn on Mauro's threads if one keeps an open mind!</p>

<p>--Lannie</p>

[alan_goldhammer]

<p ><a href="../photodb/user?user_id=2381463">Rishi Sanyal</a> <a href="../member-status-icons"><img title="Frequent poster" src="http://static.photo.net/v3graphics/member-status-icons/2rolls.gif" alt="" /></a>, Mar 17, 2009; 04:31 p.m.</p>

 

 

<blockquote>

<p>So why not write a scanning tutorial for us?<br>

Because my damn unrelated Ph.D [which is quite interesting in and of itself] keeps getting in the way :)<br>

Seriously, though, I very well may, as a compendium on high quality scanning that incorporates some of the immense information detailed even here on photo.net simply does not exist (don't even get me started on <a rel="nofollow" href="http://www.amazon.com/Scanning-Negatives-Slides-Digitizing-Photographic/dp/193395230X/ref=pd_bbs_sr_1?ie=UTF8&s=books&qid=1237320994&sr=8-1" target="_blank">this book</a> ; with comments such as "the higher the density of the slide, the brighter & more contrast it will have" even in its second edition, it just makes me irate reading it).</p>

</blockquote>

 

<p>I bought a Nikon Coolscan 5000 several months ago to begin digitizing 30 years worth of 35mm slides and negatives. The Nikon manual was essentially useless and I did purchase the book linked above. It wasn't quite useless but came close. A good tutorial would be valued by all of us amateurs out here who are trying to do some good things with our negatives. I have some great scans and some that are not so great but it's largely trial and error. If your PhD dissertation is in biochemistry (as that's what mine is in), I'm happy to lend a helping hand in return for a good tutorial!!! :-)</p>

[rishij]

<blockquote>

 

<p>Rishi, even the color tones are better! Holy cow! What one can learn on Mauro's threads if one keeps an open mind!<br>

--Lannie</p>

 

</blockquote>

<p>Hopefully, Lannie, you're joking :)</p>

<p>I went ahead and post-processed the latter file more thoroughly, hence the color/tone difference. I should probably re-post with both tuned the same, argh. Just bein' lazy.</p>

[robert lee]

<p>PhotoAcute works nicely but the resolution gain, for the current implementation at least, isn't actually all that great. At some small fraction of the resulting linear doubling of image dimensions, the improvement certainly isn't enough for really large prints. </p>

<p>This is why stitched composites can really shine. As for capture speed, it's about 3sec per frame with a Gigapan head. Figure on having to shoot 6 frames to get a quadrupling of native camera resolution.</p>

<p>By the way, it really isn't an either/or situation. The tonality advantage and noise reduction aspects of PhotoAcute can be had by a contrast blending step, vis. the open source enfuse package. Stitch the blended frames for a resolution increase. It's not as hard as it sounds - Hugin will jointly do both (and more actually.)</p>

<p>I'll include an example of what to expect from just enfuse in isolation. The individual frames were shot with a Canon A720IS compact digicam at 800ISO, and in JPEG. This is no DSLR and profound noise at these settings is to be expected.</p><div></div>

[robert lee]

<p>Crop of one of the individual frames, straight JPEG out of the camera.</p><div></div>

[robert lee]

<p>Now the blend.</p><div></div>

[robert lee]

<p>Another set of crops.</p><div></div>

[robert lee]

<p>Here's the second blend.</p><div></div>

[rishij]

<p>Alan - ha! It *is* in fact... well, medicinal chemistry technically, but mostly biochem/biophys -- I'm working on understanding the packaging motor of viruses...</p>

<p>I'll make a deal with you: help me purify my intractable viral protein & I'll take some time out to write a book :)</p>

<p>The sad thing is, by the time I (we, because certainly it'd be a joint effort with all the help & work provided by folks like Roger Smith, Erik de G, Mauro, Mendel Leisk, Scott Turner, Robert Budding... countless wonderful people in these threads have contributed to my knowledge & efforts) ever get around to writing one, film will probably be dead.</p>

<p>I'll try anyhow, for the sake of posterity. But first I'm trying to finish my prototype holders for the Nikon LS-4000/5000 and the Minolta DSE 5400 aimed at holding film as flat as wet-mounting would allow. Still not sure if the concomitant light source modification required for the LS-4000/5000 would allow ICE functionality though... so I've got some kinks to work out. Famous last words.</p>

<p>-Rishi</p>

[rishij]

<blockquote>

<p>PhotoAcute works nicely but the resolution gain, for the current implementation at least, isn't actually all that great. At some small fraction of the resulting linear doubling of image dimensions, the improvement certainly isn't enough for really large prints.</p>

</blockquote>

<p>Robert, have you tried different profiles when processing the super-resolution image? I've found that using profiles for higher-end cameras results in better resulting resolution, but sometimes less noise removal. Not sure why, as I don't understand (yet) how the different profiles for the different cameras actually function.</p>

<p>-Rishi</p>

[Landrum Kelly]

<blockquote>

<p>Hopefully, Lannie, you're joking :) --Rishi</p>

</blockquote>

<p>No, Rishi, I am just that stupid. I have been looking at all the pairs of files that had different tones and wondered exactly why they varied so much. (Does film v. digital routinely affect tonal accuracy that much?)</p>

<p>--Lannie</p>

<blockquote>

 

<br /></blockquote>

[swilson]

<p>

<p>Rishi, I am trying PhotoAcute and only seeing very small improvements. I am thinking in your test case one of your shots was out of focus or motion blur and the large gain you saw was mainly from PhotoAcute picking detail from one of the sharper images.<br>

I also find it much slower then stitching.<br>

Still it is an interesting product, I will play with it more and see if I can’t get better result from it.</p>

</p>

[rishij]

<p>Still can't tell if you're being serious or not, Lannie, but for whatever it's worth:</p>

<p>Scanned negative film has no color accuracy, so to speak. The scanning operator + software determines the overall color balance. Of course, there are red, green, and blue sensitive layers in the film, so the color information is accurately recorded (unlike Bayer sensors, hence why the blue rivers turned black in the Nikon digital capture). However, what those colors map to in the final digital file is up to the scanner/software/operator.</p>

<p>In slide film, the resultant colors are whatever the film manufacturer specified the colors to be given E-6 processing. The color dyes in both negative & slide film respond to a wide range of colors, hence the color information is vast in comparison to a Bayer sensors who's color filters only record narrow bandwidths of the spectrum. What that color information maps to in the final chrome is up to the film manufacturer. For Velvia, arguably, it's quite pleasant, not necessarily accurate.</p>

<p>The nice thing about digital RAW capture is that by shooting color cards in real world shooting scenarios yourself, you can make your own profiles (using Adobe DNG Profile Editor) that accuratel reflect the color response of your digital sensor in the real-world. That way, when you use said profile to process your RAW file at home, you can actually achieve the colors that you saw out there in the real world when you shot the image (provided you have a profiled monitor of course).</p>

<p>This is quite powerful.</p>

<p>It's also possible with film, just arguably much harder. Especially since, in my observation, Velvia colors for the same subject vary wildly depending on exposure value (EV). So you'd have to shoot color cards for different EVs... Then use some open source profiling software that works with Gretagmacbeth color charts, then take the raw scan file and apply the profile to it. I dunno how well that works, honestly.</p>

<p>What I do is just profile the film+scanner combination, then apply that profile, getting me exactly the colors I see on the film on a 5000K lightbox. It's quite wonderful, actually, since Velvia already produces very pleasing colors. No post-processing necessary -- just apply the profile. Velvia chemistry does all the post-processing :)</p>

<p>-Rishi</p>

[rishij]

<blockquote>

<p>Rishi, I am trying PhotoAcute and only seeing very small improvements. I am thinking in your test case one of your shots was out of focus or motion blur and the large gain you saw was mainly from PhotoAcute picking detail from one of the sharper images.</p>

</blockquote>

<p>Scott, I also initially thought the same thing. But, no, the original image you see there is the sharpest of the 3! No motion blur... it was tripod mounted with a 2 sec shutter delay. No shutter vibration b/c, well, technically it's a P&S :)</p>

<p>Try the different profiles. For the LX3 I use the Leica M8 profile. The Canon Rebel profiles give you pretty good resolution jumps, but less noise reduction. Again, not sure why. Try one of the SLR profiles even if you have a P&S. Also, I was working with RAW images from the LX3, not JPEG. Dunno how much of a difference that makes. PhotoAcute uses Adobe DNG converter to process the RAW files.</p>

<p>-Rishi</p>

[rishij]

<p>Scott & anyone else, here try yourself with my 3 original RAW files:<br>

<a href="http://staff.washington.edu/rjsanyal/Photography/LX3/">http://staff.washington.edu/rjsanyal/Photography/LX3/</a></p>

<p>Here you can download a trial of PhotoAcute:<br>

<a href="http://photoacute.com/">http://photoacute.com/</a></p>

<p>Try using the Leica M8 profile when doing the super-resolution stitch.</p>

<p>-Rishi</p>

[bernardwest]

Rishi & Robert....<p>

 

Wow... I'm interested in these bits of software you are talking about here. Definitely some nice results for both of you. Robert, I'm not sure of the workflow you are using here. Is this blending say three bracketed exposures, or is it doing a pixel binning type thing like Rishi's Acute? If it is the former, I almost hate to mention this here, but a guy over on Luminous Landscape (ouch...) forums has written a fantastic piece of software which can blend bracketed <b>raw</b> exposures. Last I heard he was working on getting it to output a blended raw file in the form of a DNG. <p>

 

Ok, back to the film world....<p>

 

p.s. Rishi, can you manage to get a bit of black and white film chemistry into the thesis ;)

[swilson]

<blockquote>

<p ><a href="../photodb/user?user_id=1742025">Roman Ponomarjov</a> <a href="../member-status-icons"></a>, Mar 17, 2009; 08:51 p.m.</p>

 

<p><em><strong>Poor scans - May explain why some switch to Digital.</strong></em><br>

May be, but it’s partially…</p>

<p>The main reasons seems to me is lack of adequate information. Can anybody imagine that such test may appear on Nikon Imaging website or in Popular Photography magazine? Obviously not. Seems like our so called “free public media” sits under tight financial control of digital manufacturers. And how many of us have seen that Nikon is promoting its scanners? Any ads, commercials, articles, announcements… Apparently Nikon already ceased the production of Coolscan V ED. And the reason is simple: it’s affordable, it delivers good performance, it’s popular and in demand. Yes, of cause, it contributes in film photography popularization and makes negative impact on digital camera sales.</p>

<p>From my personal experience, 4 years ago I did not believe that scanning my produce good result. The only I’ve read and seen in our magazines it’s how a cheap digital camera can easily overpass scanned film in quality, gray range, res etc. and I was thinking quite serious about buying a DSLR. Until I took a look at Les Sarile Film 2 posting. This posting had drastically changed my mind. Since that I purchased my Coolscan V and never look back at digital again.</p>

<p>Seems to me that Les and Mauro do for film photography popularization much more than the leading film manufacturers do. They did lots for all of us on this forum to keep our spirit alive. Guys, I guess the digital camera makers must hate you….</p>

 

</blockquote>

<p>If you believe that Nikon was worried about the Coolscan V ED reducing digital camera sales you are rather out of touch with what has been going on the last 10 years or so. <br>

 

<p>Nikon and other camera manufactures are concentrating on the digital market because there is virtually no sales of new film cameras.<br>

 

<p>Don’t expect to see much in the way of new film scanners, for the most part those of us that were ever going to scan film have already bought ours and have moved on to other things.</p>

<p>

<p>There is no company or publication that has been out to get film. The decline in the use of film is not because of some conspiracy to do away with it, it is in decline because the public is voting with their pocketbooks.</p>

<p>You might get a few people here and there to stick with film, at least for a few more years, but this will do little with the mass movement over to digital.</p>

<p>I really don’t see this as a war between film and digital as some seem to. In the early part of this decade the digital camera manufactures were mostly trying to get people to switch from shooting film to digital, but now they are mostly trying to get people to upgrade from older digital cameras to newer ones.</p>

</p>

</p>

 

<p>

 

</p>

</p>

 

[swilson]

<blockquote>

<p ><a href="../photodb/user?user_id=2381463">Rishi Sanyal</a> <a href="../member-status-icons"><img title="Frequent poster" src="http://static.photo.net/v3graphics/member-status-icons/2rolls.gif" alt="" /></a>, Mar 17, 2009; 10:23 p.m.</p>

 

<p>Scott & anyone else, here try yourself with my 3 original RAW files:<br /><a rel="nofollow" href="http://staff.washington.edu/rjsanyal/Photography/LX3/" target="_blank">http://staff.washington.edu/rjsanyal/Photography/LX3/</a><br>

Here you can download a trial of PhotoAcute:<br /><a rel="nofollow" href="http://photoacute.com/" target="_blank">http://photoacute.com/</a><br>

Try using the Leica M8 profile when doing the super-resolution stitch.</p>

 

</blockquote>

<p>Hi Rishi,</p>

<p>

<p>I don't seem to have a raw converter that will read those files, can you post high quality jpgs of them, or even tiffs?</p>

</p>

 

[swilson]

<p>Hello Rishi,</p>

<p>So I converter to DNG and PhotoAcute loaded them and ran fine, and I see very much what you saw. <br>

 

<p>But I am concerned that perhaps the gain would not be so great if they were working with well converted raw files to begin with. Raw files with no sharpening tend to look soft, so this may be more a case of sharpening then combining photos. How sharp of an image can you get from one image when you convert from raw to jpeg?</p>

</p>

[Landrum Kelly]

<blockquote>

<p>Still can't tell if you're being serious or not, Lannie. . . . --Rishi</p>

</blockquote>

<p>No, Rishi, I really am just that ignorant.</p>

<p>Seriously, your comments were very helpful. I tend to reinvent the wheel every time I start processing a RAW file. It's high time I learned something about profiles.</p>

<p>Thanks, Rishi. You're an inspiration. (No kidding.)</p>

<p>--Lannie</p>

[Landrum Kelly]

<p>That's a cloth-bound book, Les. The texture is real.</p>

<p>--Lannie</p>

[rishij]

<blockquote>

<p>What texture from the book? The only texture there is JPEG artifacting?</p>

</blockquote>

<p>Haha, get your eyes checked, Les.</p>

<p>Scott, thanks for reminding me I need to pick up a copy of Horowitz & Hill :)</p>

<p>Also, Scott, perhaps you're right -- perhaps the RAW converter for the LX3 is still in its neophytic stages? I don't like to compare to JPEGs out of the camera b/c camera JPEGs are, even at the highest setting, utter crap (for the LX3, that is). Too much artifacting.</p>

<p>My comparison has sharpening only set to '45' on a scale of 100 in LR, so, sure sharpening could be increased. Basically I feel like PhotoAcute is allowing LX3 images to be on par with a higher end dSLR. The question remains what it'd do with actual dSLR images. I don't have one, but I'm sure one of you guys on this thread could give it a shot. Ahem, <strong>Bernie & his 5D</strong> :) Else I'll try & borrow a Rebel XTi over the weekend from a friend.</p>

<p>Lannie, you're awfully nice :) Color profiling is amazing, if not a world of hurt. Try the open-source LPROF, but you may or may not have a nightmare trying to compile it. Once you do that you'll have to solve the 'blacker-than-black' problem, something I still haven't solved but have an idea of how to address (so do the devs of LPROF, just haven't gotten around to chatting with them again about it to see if there have been any updates). Barring working with open-source packages, you'll have to pay out the wazoo for a pre-compiled, maintained package like basICColor (they've fixed the blacker-than-black problem, btw). Monaco sells a package too, I'm just scratchy on the details b/c I've been sidetracked, temporarily, from color-management issues (which are much more of a solvable problem than, say, film flatness... hence why I've been worrying much more about the latter as of late).</p>

<p>Have fun, Lannie!</p>

<p>Cheers,<br /> Rishi</p>

[swilson]

<p>Les here is a hint, jpeg artifacts don't tend to run at angles, that is the threads of the binding you are looking at there.</p>

<p>

<p>But just for you here is a tiff straight from the raw file, hard for there to be jpeg artifacts in that.</p>

 

</p>

[c._sharon]

<blockquote>

<p>Les - The first obvious visual clue is how the 5D is framed</p>

</blockquote>

<p>This has already been answered.</p>

<blockquote>

<p>The mottled appearance of the scans is something I have never encountered in such fine film either.</p>

</blockquote>

<p>Are you saying a pro lab does not know how to use a scanner? Possible, but unlikely.</p>

<blockquote>

<p>Having repeatable independently verifiable tests means they can cook up anything they want but doesn't mean they will get away with hiding the truth.</p>

</blockquote>

<p>As I said before, Google is your friend. Here is another link to a 1DsII v Mamiya 645 test. 1DsII does better. Now this is not the same set up as in the previous link but is another FF dSLR vs MF film. You can also find links to other tests between D2x and film. If a APS dSLR like D2X can equal 6x6 film and give drum scanned 6x7 film a run for its money even if not quite matching or surpassing it, a FF dSLR like 5D surpassing 6x6 film is not implausible.<br>

http://photo.nemergut.com/equipment/canon1ds/markii.html<br>

http://www.d7x.com/HTML/V13/D2xVs6x7.html</p>

<p>So referring back to the original premise of this thread, poor quality scans are not the problem.</p>

[swilson]

<blockquote>

<p ><a href="http://www.photo.net/photodb/user?user_id=2381463">Rishi Sanyal</a> <a href="http://www.photo.net/member-status-icons"><img title="Frequent poster" src="http://static.photo.net/v3graphics/member-status-icons/2rolls.gif" alt="" /></a>, Mar 17, 2009; 11:34 p.m.</p>

<p >Scott, thanks for reminding me I need to pick up a copy of Horowitz & Hill :)</p>

</blockquote>

<p >The Horowitz & Hill book is great, I am on my thrid copy, long story. They have a very nice treatment of low noise circuits and good coverage of transistors and noise. I am a bit of a transistor junky, I love that a $0.05 part can do what an op amp would take $5.00 to do. </p>

<p > <br>

 

<p>If you want to see the rather eclectic collection you can see the whole book case here, a rather large file that might not show in a browser window.</p>

<p><a href="http://sewcon.com/samples/books.jpg">http://sewcon.com/samples/books.jpg</a></p>

</p>