Epson 3170 - blurry film scans, what's going on?

[nang]

I recently bought an Epson 3170 as I need both a film scanner with MF

capability and the ability to scan documents. I just got some velvia

slides back today (mounted) and I've scanned a few, but I'm really

disappointed in the quality of the scan. I've viewed the slides

under magnification and they are razor sharp, significantly sharper

than what I'm getting in my scans.

 

I've (with hope) successfully attached images to illustrate my

point. This is using the normal Epson scan drivers. I've tried

Vuescan and it was even worse, although I used multiple passes in

Veuscan, if that has any effect.

 

Is there some way to improve the sharpness of my scans? I'm using

professional mode in the Epson scan software.<div></div>

[nang]

Here is a closeup of one of the buildings, scanned at 3200 dpi.<div></div>

[Ed_Ingold]

The scan and detail look about typical for MF on an Epson flatbed. I have a 2450. I believe the 3170 is comparable. While it is a great improvement over previous flatbeds, the results are disappointing in comparison with a dedicated film scanner.

 

The claim of 3200 dpi resolution is puffery. On a good day, with care, you can get about 1000-1200 dpi resolution. In any event, you need a lot of sharpening to get reasonable results. Medium format on a flatbed scanner in this generation do not look as sharp as 35mm scans on an LS-4000, despite the format size. I have taken large group photos with a D1x that have better facial clarity.

 

From my personal experience, the only reason I still have an Hasselblad system is the results from an LS-8000, having suffered the Epson flatbed route.

[kenneth_beatty]

I know this is not the solution you were looking for but I had to play with it to see what I could do by useing a smart sharpen and a high pass sharpen. It is just a band aide so my suggestion would be to look into a dedicated film scanner.

[steven_clark]

Learn to use unsharp mask,and use it well. Personally I recommend multiple passes of different radiuses, EG one to clear up the fogginess, one to enhance edges, one to enhance detail, or something similar. really this is the kind of situation where you'd want a perfectly tuned deconvolution to get the most out of the resolution that's there, but it's not really practical to find a good one by trial and error.

[steven_clark]

Well, it looks like there is an <a href="http://math.nist.gov/mcsd/highlights/blinddeconvolution.html">answer</a> after all. I don't think anybody's come up with a good consumer version yet. Maybe I'll have to look up a copy of those papers and see what I can do with it.

[jem]

Thats a March 2000 link. And I suspect it is one of those things that is more interesting in theory than practise...I suspect in real world scenarios and under conditions that aren't controlled, it would fall down. And I suspect these things always will - it strikes me as fantasy to suggest that you can replace information that is simply not there (a la the magic buttons in tv shows). If you're talking about using an algorithm to reverse certain controlled types of blur, then maybe you can get OK results, but not results suitable for fine art printing.

 

Given the enourmous volumes of money in image processing and related software, if this was applicable I'd expect it to already be out there by now...it doesn't take four years to move a practical algorithm to market in the digital imaging field.

 

Then again, maybe it will work and will be great...but in the meantime, the Epson flatbeds are ok for the money but lets face it if you're into producing quality work, they just aren't good enough.

[pengland]

Nick, I use an Epson 2450 to scan B&W negs. For the best results on this set up I do not use the neg carrier supplied with it but instead sandwich the film between the scanner glass and the anti-Newtoned neg carrier glass from my enlarger. I scan at 2400. You would have to demount a slide to do this properly. Give it a try with plain glass if you cannot get your hands on some anti-Newtoned stuff. Ground glass might do the trick. Phillip.

[jbq]

That's comparable to what I've seen with my 3200.

 

On the 3200 I never bother to scan at anything higher than 2400 dpi (I see extremely little difference between 1800 and 2400 dpi), and even that takes some heavy sharpening. This is acceptable for me with toy MF cameras and my LF gear with its cheap lenses, but for 35mm I upgraded to a Minolta Dual IV and it is just immensely better.

[custom film holders for fl]

Using Vuescan (or most any program) with the multiple pass option engaged often results in flatbed scans that are less sharp than single pass scans. The main reason is that when the scanner head travels back to the starting point to start the second scan, many times it can be off by just a pixel or two compared to the original starting point. This results in softness once the two scans are combined.

<p> Doug<p>

<a href="http://home.earthlink.net/~dougfisher/holder/mainintro.html">Dougs

MF Film Holder for batch scanning of 120/220 medium format film with flatbeds</a>

[steven_clark]

Actually in theory it should be possible. Minus pixels that might actually be linear combinations of their neighbors a perfect deconvolution will pull out just as many perfectly sharpened pixels as it takes in blurred pixels. The idea is that the data is there, it's just each pixel's data is spread out into the surrounding pixels. If you know exactly what that blur is, you can mathematically reverse it. Unsharp masking is just an estimated kluge in comparison, namely further blur the image and subtract that blur from the already blurred original image. Actually, it kinda surprises me that Epson hasn't tried to come up with a pre-created deconvolution matrix just for each scanner, 'cause done right it would probably improve their reputation immensely.

[lachaine]

You know, I've gotten better results than yours with just a Canoscan 8000F 2400dpi flatbed from 35mm slides and negatives. And, I've seen a great many nice sharp scans of medium format film done on Epson printers. There must be something not right about your scan.

[madwand]

Focus Magic does do decent software sharpening, but it's not fully blind. It already costs so much in time that you'll want to use it only when necessary. Sometimes USM gives comparable results in far less time. But for the tough cases, esp. shake for example, Focus Magic can't be matched.<p>

 

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

 

There's an example of it in practice deep in this thread.<p>

 

<a href="http://www.photo.net/bboard/q-and-a-fetch-msg?msg_id=008HwB">http://www.photo.net/bboard/q-and-a-fetch-msg?msg_id=008HwB</a><p>

 

When I used a 2450, I found that sharpening was routinely needed (and USM was typically fine. <a href="http://www.scantips.com/simple6.html">http://www.scantips.com/simple6.html</a>). But I think that your case is somewhat worse that those I'd seen, and in your place I'd look further into adjusting the distance / identifying the focus point / having the scanner replaced / adjusted if necessary.