Was this long epistle supposed to make me read for a sufficiently long
period of time to forget that you didn't actually answer my question
about 5 pass sharpening? It's a good thing you're not a talking head
else I'd think you were deliberately avoiding it. :-)
But I agree with your assessment of 10-12 MP max resolution for a 35mm
film scan. That's roughly equivalent to 60 lines/mm which is about as
good as you can get from tripod based photos in the field regardless of
film and lens resolution. Provia 100F is capable of more than that but
the camera would probably have to be mounted on a steam engine... inside
your studio. And we know OMs very much need that unless the mirror and
aperture are pre-fired.
How about illustrating exactly what you're talking about here. Can we
see a small 1:1 pixel crop of a single scan and then the same from your
multi-pass, multi-sharpened image. I'd like to actually see the
difference in detail.
Chuck Norcutt
On 6/12/2012 10:54 AM, Ken Norton wrote:
>> Care to describe your "5 pass sharpening process"?
>
> The act of scanning is actually that of taking a picture of a piece of
> polyester that happens to have some form of image on it. The scanner
> is a digital camera with a lens.
>
> In a single pass scan, this is a single picture. The problem we run
> into with single pass scans is aliasing artifacts where grain (or dye
> clouds) are sampled in a such a way that we are likely to get wild
> swings in the values of each pixel. So, for the highest level of
> quality we overscan. Each time the scanner photographs the image,
> there is a very tiny offset as the stepping motors don't quite get
> things exactly aligned. With this multi-pass scan, we effectively
> increase the diameter of the sampled area for each pixel. This reduces
> the grain aliasing.
>
> Unfortunately, overscanning like this tends to reduce sharpness. I
> have found that there is no specific setting you can use to bring the
> image back to maximum sharpness. For one thing, this is so film
> dependant (and scanner dependant), that the settings which work for
> the original Provia 100 won't work for Provia 100F. Velvia 50, 100 and
> 100F all scan differently too. Let's not even talk about print films.
>
> So, overscanning reduces sharpness. OK. But does that mean that the
> information is gone? Of course not. This is where the imaging software
> as used by CSI comes in. I only need six pixels to be able to recover
> an entire page from a newspaper. ;)
>
> Not really.
>
> Basically, what we have to do is apply a deconvulution style recovery
> of the image. This may involve multiple passes of USM with varying
> diameters. If you apply the diameters in the wrong order you will end
> up with halos and other crunchies. One thing I very much appreciate
> about PWP is the sharpenening algorithms which are a bit better than
> what Adobe uses. OK, quite a bit better...
>
> I find that with a massive overscan (5-10 passes), I need about five
> passes of sharpening. WIth a single or dual scan (HDR style scan), I
> can only do two passes before things get nasty.
>
> The old film vs. digital comparisons assumed a simplistic approach to
> scanning and detail recovery. I am, by no means, an expert on these
> things, but I'm definitely able to recover a whole lot more detail
> than what the talking heads have claimed possible.
>
> One little trick of the trade about scanned film vs. pure digital
> imaging. Using a very good resize algorithm (again, like the one in
> PWP), I can downsize my 20 megapixel equivalent scan to a 10 megapixel
> equivalent image, apply another round of post-scaling sharpening
> (again, using the awesome sharpening algorithm in PWP) and end up with
> an image that may be superior to a digital camera image of similar
> size. As we all know, a 10 megapixel camera is capable of some pretty
> incredible stuff. Or has Bob been yanking our chains with his Maine
> pictures all along? Once reduced in size and sufficiently cleaned and
> sharpened, the image can then be upsized exactly in the same manner as
> if it originated in a digital camera in the first place.
>
> Now, let's talk about equivalents? Is a 20MP digital camera going to
> outperform a 20MP scan? Of course! I figure that an idealized scan of
> a decent slide film is going to come in at about 10-12MP. However,
> those 10-12MP are per-pixel sharp with no aliasing. A sharp edge is a
> sharp edge, not a two or three pixel transition. (Or in the case where
> two opposite colors come together like red and green, you don't get a
> gray line inbetween).
>
> Unfortunately, we now run into other limits: Provia has the lateral
> halation problem which tops out maximum usuable resolution. Lens
> handling techniques, which result in defraction blurring. And then the
> dreaded OM system aperture actuator vibration. It's a rare non-flash
> image that will maximize a 4000 dpi scan. Flash? Sure. Even the Gary
> Reese Lens Tests reveal how difficult it was to maximize sharpness.
>
> When do I use multi-pass overscanning? Not for every shot. In fact, I
> usually only use it for something that I know will end up printed big
> or when there is so much dynamic range which needs to be protected or
> recovered. The churn and burn stuff is usually single-pass. The latest
> versions of Kodak Portra do just fine with single-pass. Slides with
> extreme dynamic range, such as "Sunburst Lake", need a lot more work.
> A LOT more work.
>
> The point of this epistle, is that with a creative and disciplined
> approach to scanning technique and image recovery, there is a whole
> lot more detail buried in these old film images than what we generally
> give the technology credit for.
>
> When I'm in the darkroom, I use a grain focuser when making a print. I
> actually have three of these grain focusers. One of them is a very
> nice one which I managed to score from Bill. With my two 'affordable"
> ones, I thought I could see and focus the grain. On a bad day, I can
> see more detail in this highly magnified image than what my scanner
> has ever been able to recover. But then with Bill's higher-end
> focuser, I saw stuff I never saw before--particularily the point when
> diffraction was kicking in with the enlarger lens! The point is that I
> see far more detail in some of the negatives than what the
> digitization process is able to reveal.
>
> So, as an engineer, I start comparing the differences and then try
> different techniques to get the two technologies to match. This is
> what led to the discovery of overscanning combined with multi-pass
> sharpening. As to being able to recover everything possible from a
> Ilford Pan-F negative? Based on what I see, it would take at least a
> 10000 dpi scanner to come close. By no means have I "arrived" with a
> final set of conclusions, or techniques. It remains a work in
> progress.
>
> To give credit where credit is due, I wouldn't have figured out this
> process without a little hint from my Dad. He is a retired toolmaker
> in the aerospace industry who spent the last dozen or so years of his
> career building prototypes and one-off devices using what was at the
> time the most accurate grinder in the world. (Probably still is, the
> latest stuff uses lasers to get even closer). His claim to fame was
> being able to actually get an additional 400% improvement in accuracy
> than what the technology could ever allow.. The techniques he
> developed became trade secrets, but I learned enough from him to be
> able to apply a similar logic to the scanning process.
>
> AG
--
_________________________________________________________________
Options: http://lists.thomasclausen.net/mailman/listinfo/olympus
Archives: http://lists.thomasclausen.net/mailman/private/olympus/
Themed Olympus Photo Exhibition: http://www.tope.nl/
|