[Leica] Re: digital quality

[Jim Brick <jimbrick@photoaccess.com>]

Thank you Erwin.

I've been endlessly preaching this information, but from a different point
of view.

It is refreshing to see it said, from your point of view, in a most
eloquent and understandable manner.

There should be little doubt in anyone's mind about the capabilities of
analog film verses the digital process.

And, as you said, digital imaging and prints do indeed have their place in
the photographic world.

Jim


At 09:13 PM 7/25/00 +0200, Erwin wrote:
>An intriguing topic, this one about the image quality of a  digital print.
>I did some densitometric measurements on digital prints, made by Epson.
>These were promotional and of very high quality. Measuring white gave
>D=0.04, which is as good as the best BW papers I have tested. But measuring
>black gave a D=1.42, which is very good for a colour neg print, but far
>below what you can get with a BW print, which can handle easily D=2.30. This
>simply tells you that he brightness range of a digital print is 1:26,
>compared to 1:199 for a BW print and 1:1000 for a slide projected on a
>screen. As reference note that the average computer or TV screen has a range
>of 1:30. 
>Whatever your opinion on digital prints, it is undeniable that the
>brightness range is much compressed in the shadow area and you will loose
>valuable information when printing from whatever scan you have.
>A second important aspect is the linearity of the CCD sensor versus the more
>natural logarithmic respone of the silver halide. A digital capture or
>recording then is always a discrete sampling and they need to use a linear
>or a logarithmic mapping to translate the analog nature of the film. A
>logarithmic mapping would be best, but the amount of data is too high so a
>compression algoritm has to be used, which always means reduction of
>information. 
>If we scan a negative with 4000 dpi, it would be logical to print with the
>same amount of dpi as a pixel to pixel mapping is needed. This is not yet
>possible. Scanning with 2800 dpi has the same logic. Now 2800 dpi delivers
>110 lines per mm, assuming that a dot represents a line. A typical ISO100
>film resolves 150 to 200 lines per mm, as does a Velvia slide film. Typical
>grain size is 2 micron, while the dot size of a 2800 dpi scan is about 10
>micron. So I cannot imagine that scanning a fine grain slide of negative
>will capture the grain size. The 200 lines of the Velvia would have a size
>of 5 micron,which cannot be capured by a scan with a minimum spot size of
>twice that area. 
>So whatever parameter you use, brightness range, image size of a small spot,
>grain size, the analog process has a much higher capacity for recording
>spatial information. The several degrading steps in the digital process and
>the inherent limits of the technique itself, deliver a lower quality product
>than does the analog technique.
>To see grain with a ccd capture device, you need a lens with a resolution in
>the 1 micron area. This can be done of course and Crossfield in a paper
>about image quality note that they use high end scanners with this
>capability. One scan however would be a 28.000 dpi (!) scan. Or a 1 micron
>scan of a 35mm negative would generate  864 million pixels which has to be
>recorded 3 times for colour info too and so would have a 3 Gigabyte file for
>one 35mm negative. Note this size is needed to capture the grain image.
>As long as we need image quality on a A4 size, on the threshold level of
>what the human eye can detect at normal looking distances, a digital print
>will suffice and even impress. But that is not the same as implying that the
>digital capture of information is in the same ball game as what silver
>halide can record. 
>
>Erwin           
>