Comments interspersed below.
At 8:00 PM +0000 12/23/02, olympus-digest wrote:
>Date: Mon, 23 Dec 2002 21:16:54 +0800
>From: "C.H.Ling" <chling@xxxxxxxxxxxxx>
>Subject: Re: [OM]: Dipping our Toe Into Digital, 16MP DC vs film again
>
>- ----- Original Message -----
>From: "Joe Gwinn" <joegwinn@xxxxxxxxx>
><snip>
> > When comparing various cameras and scanners, we probably should always
>ascertain the optical resolution of the green channel alone, and use that as
>our comparison. These numbers will be directly comparable to the
>resolutions of black&white film and cameras.
> >
>
>Well, thanks for the detailed analysis. I think it is not so proper to count
>just the green channel, assume you use the CCD for B/W photo, every pixel
>will be used for resolution information. For color it will also be close,
>but some details' color could be wrong, making a not so "clean" picture but
>the "resolution" should be there.
I don't think I explained my point well enough, probably because I tried to use
too few words. Here are more better words.
In a digital camera with tricolor pixels, the spatial resolution is set by the
color with the most pixels, and the other colors do not increase the spatial
resolution, only the color resolution. So, to estimate the true spatial
resolution of a camera with a RGB ratio of 1:2:1, one uses the green pixels
alone. The other two colors, red and blue, will have half the resolution of
green.
A black&white only digital camera has unfiltered pixels, so all pixels
contribute to spatial resolution.
A color camera being used to produce black&white pictures cannot have better
resolution than the color with the greatest resolution, typically green.
Because digital cameras are either 1:1:1 or 1:2:1, green is always a good
choice for determining spatial resolution. So, I use the green channel, and
also compare it directly with B&W only pictures as well.
>It is very difficult to make a "true" resolution assessment here. Just like
>the resolution of film camera system, as I mentioned before, a CCD pixel
>will contain 256 different levels (assume 8 bits) but the film color
>(information) is composed by grain, a few dots may be good enough to compose
>a line to represent resolution data but this few dots will not be as
>informative as a CCD pixel. This favor to film, in resolution test film will
>be the winner.
>
>Since each CCD pixel is so informative and clean (compared to the dot grain
>of film), that is why in many case a 4-6MP DC showing picture quality better
>than film. Ok, the details may not be as good as film but overall it look
>better.
Yes, it will be some time before digital can beat film in resolution. Where
digital already exceeds film is reportedly in dynamic range, linearity, and
noise, at least in the expensive cameras. Not that 8 bits is enough, though.
The top grade scanners use 12 bits or more per color, and the astronomers use
16 bits per color.
The linearity is probably why good digital cameras are reported to handle skin,
sparkles, eyelights, et al, so well.
> >
> > Actually, this leads to a good question for CH: Has he or can he measure
>the resolution of his scanners? If one can find a suitable resolution chart
>and scan it, it ought to be easy. The scanner maker probably also knows,
>but may not be willing to publish the modulation transfer function.
> >
>
>I don't have a professional slide for scanner test, but I do have a lens
>test slide, below is a crop of Zuiko 50/3.5 at F5.6 shot, exposed with flash
>so vibration should be minimized. The size is 5.2mm x 2.1mm, download it in
>PhotoShop to check the resolution, to me it looks over 75lp/mm.
>
>http://www.accura.com.hk/50-02.jpg (200K)
I looked at the picture, but I don't know how to interpret it. What is the
purpose of a lens test slide? To test slide duplicators?
>Below is a message I posted before to compare a simulated 16MP DC and 35mm
>film system, enjoy!
>
>==============
>May be we start another interesting topic about how many pixel a DC can be
>comparable to a common film we are using.
>
>I just perform a test with E-10, shooting at 140mm (35mm equ.) and Fuji
>Provia F shooting with 35-105 at 70mm.
>
>E-10 is a 4MP DC. I try to simulate a 16MP DC by setting it to 140mm and
>compare it to a shot taking at 70mm. The file is a bit large (around 1.1MB
>and 1.7MB each). But it is interesting to have a look.
This ought to be reasonably valid, if the optics in the E-10 are about the same
quality as the 35-105 lens.
>The 70mm shot taking with Zuiko 35-105mm (early version) at 70mm F8, tripod
>and shutter speed was 1/250. Scanned at 4000dpi and cropped the center
>17.9mmx13.4mm.
The scanner is operating at a claimed 4000/25.4= 157.48 pixels per millimeter,
so the kept part of the scan will be (17.9)(157.48)=2818.89 by (13.4)(157.48)=
2110.24, call it 2829x2110.
>I have made three shoots, two with 35-70/3.6 and one with
>35-105, they all have similar resolution but the 35-105 has slightly higher
>contrast.
>
>http://www.accura.com.hk/70mm.jpg
>
>This is a E-10 shot at 140mm, ISO80, F4.8, 1/640s, sharpness setting: low
>(F4 is the best aperture of E10)
>
>http://www.accura.com.hk/E10-140mm.jpg
>
>I don't make any comment, download them and check it out yourself by making
>side to side comparison, you may need to see the different in 200% view.
To my eye, the E-10 shot is nicer-looking, but the scanned film has higher
resolution. Sort of as expected. So, there you have it.
If you have the energy, could you make test shots like these, only with wider
dynamic range, and including some specular highlights? Thanks. The claim is
that digital does better than film in such situations.
Joe Gwinn
< This message was delivered via the Olympus Mailing List >
< For questions, mailto:owner-olympus@xxxxxxxxxxxxxxx >
< Web Page: http://Zuiko.sls.bc.ca/swright/olympuslist.html >
|
