On 3/2/2011 7:54 AM, Sawyer, Edward wrote:
> The best MP ever tested that I know of was the Questar 700mm. It was head and
> shoulders above all the others, incl. Zuiko.
Yup, sometimes you get what you pay for. It is also head and shoulders above
the rest in price.
> BTW, contrast is vitally important in terms of lens specs, regardless of
> scanning vs. film.
Here, I disagree. Useful, sure. Vital? Well that depends on a lot of things,
including what other lens performance
factor it is a trade-off for, subject, film/sensor, intended look of the final
output, etc.
> Contrast and resolution are bascially the same thing, measured at different
> frequencies. Higher contrast is always better.
Here, I really disagree. I don't care about the theory, assuming we are talking
MTF, for two reasons. First, I don't
understand it. I think I may have understood it for a few minutes at a time a
few isolated times; but certainly not for
more than maybe 30 minutes, combined total. Maybe it's the name? Maybe I'm just
dumb. I did manage good grades and
placing in a couple of honors courses as a physics major at Berkeley a long
time ago.
Second, and more importantly, it just doesn't agree with my experience. I've
worked on thousands of images in my digital
darkroom. Recovering softness, lack of resolution of detail, is always more
difficult and less successful than
recovering a full dynamic range from low contrast.
Third, it just doesn't make sense at a very simple, practical theoretical
level. Imagine a subject like classic
resolution targets, fine alternating lines of pure black and white with near
100% contrast at their edges.
Now imagine a lens or process that reduces the contrast between black and
white, while maintaining the perfectly sharp
edges between them. Even if the difference is only one digital value, say 0 and
1, or 8,000 and 8,001 or 65,000 and
65,001, the original may be perfectly recreated from the processed copy. In
more practical terms, imagine a 16 bit
grayscale image that occupies only about a quarter of the histogram. In the
middle, that represents a luminance range of
960 values, almost four times the number in an 8 bit JPEG. With smarter
exposure, in scanning or digital camera, but
still short of clipping, it represents a luminance range of 30-60,000 values.
With that kind of range, it's easy to stretch and interpolate to a very nice,
full range histogram and full tonal range
image. Certainly more than can be represented in a web image, and I imagine,
more than any human eye can differentiate
in a print. Whether film has such an ability to differentiate all those fine
tonal values and which scanners may do so
accurately, I don't know. But that's a problem with the medium, not the theory,
and is one reason I posited that it DID
matter with film. With decent digicams, though, the whole range is captured in
full detail.
This contrast manipulation is illustrated on the first line of boxes.
<http://www.moosemystic.net/Gallery/tech/Contrast_Resolution/Contrast_Rez.htm>
I've used a pattern, rather than a photographic image, so the differences will
be clear. I've used one complex enough to
present all sorts of differences in line and space widths:
1. Original pattern
2. Brightness brought down to move into the middle of the histogram where there
are fewer tonal values available.
Contrast reduced until the pattern is just barely visible.
3. LCE to assure sharp contrast at edges.
4. Levels to pull tonal range to top and bottom of histogram.
Pretty hard to tell the difference between original and recovered, no?
Now imagine a lens or process that renders the subject as an image with a good
range of contrast, but poorly defined
edges. The center of the black lines is at least close to black, but there is
a gradient from center to edge such that
the edge doesn't vary in tonal value much, or perhaps at all, from the
similarly configured white line. Imagine further
that the blurring is the result of several, complex functions or lens
aberrations, not some simple, function that can
be accurately deconvoluted.
How do you recover from that? The answer is that you can usually do some good,
but really, unless the effect is pretty
mild, the detail lost isn't fully recoverable. USM, used in various ways, can
create a 'look' that's a lot like original
high resolution, but not quite. And if you look closely, you will see that the
way it created the appearance of
sharpness isn't really a recovery of what was lost.
This resolution manipulation is illustrated on the 2nd. and 3rd. lines of boxes.
1. Blur using a combination of PS Lens Blur and Surface Blur tools, so simple
Gaussian deconvolution won't work well.
2. Levels to recover lost contrast.
3. PS Smart Sharpen. Not surprisingly, considering how the image was blurred,
the Lens Blur option worked best.
4. Focus Magic, Out Of Focus option, 2 pixels.
5. Focus Magic, Out Of Focus option, 3 pixels.
- I tried other sharpening options, but none was better than these.
- Notice how the spaces between close lines have filled up with middle gray
tones.
I hope you can see that, if this were a 100% sample of a large image, these
tools would make downsampled versions look
quite a bit sharper. I've tried it, and it's true, but I'm not making another
multiple roll-over tonight. :-)
Generalized deconvolution tools can recover more or less detail, depending on
subject and what the actual lens
aberrations were. But again, it's not the same. You can fool the eye in all
sorts of ways in a smaller image, but if you
want to display big, you want real resolution from the outset.
So for me, given other things equal, I'll take more resolution and less
contrast over the reverse. If it weren't a
trade-off, I'd of course take both. But it is generally a trade-off, under the
optical "No Free Lunch" rule. There is,
of course that other trade-off, where one may buy a more optimal combination of
both for a great deal more money. Cases
in point, some of the Leitz and Zeiss lenses and the Questars
I understand why lens makers aren't leaning this way. The vast majority of
people taking pics, including many on this
list, judge a captured image on what it looks like on the LCD or directly out
of the camera. Optimizing lenses for
resolution at the cost of a flat looking original image ain't gonna sell many
lenses, even if it allows higher
resolution post processed images.
> Jacking up contrast after scanning is not the same thing as using a
> higher-contrast lens to start with.
I dunno. In the vast majority of cases I've dealt with, it pretty much looks
the same. With real world photographs,
contrast alone certainly doesn't do the job, LCE is required, too. that's the
big reason all the RAW converters have
sprouted what amount to LCE sliders under a variety of names.
Compare & Contrast Moose
--
_________________________________________________________________
Options: http://lists.thomasclausen.net/mailman/listinfo/olympus
Archives: http://lists.thomasclausen.net/mailman/private/olympus/
Themed Olympus Photo Exhibition: http://www.tope.nl/
|