[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]
Re: [APML] Fw: images
On Wed, 23 Jun 2004, Chuck Vaughn wrote:
> It amazes me how much confusion there is over bit depth, dynamic range,
> and S/N. In practicality, the finest grain films are limited to about 5
> - 6 bits and most color films are about 4 - 5 bits.
>
> Let me provide an example as to what I mean. The following image is a
> section of a Tech Pan photo originally scanned at 3048 dpi and 16 bits
> with no processing at the scanner.
>
> http://www.aa6g.org/Images/m81.psd (2.5MB)
>
> It was contrast stretched in Photoshop at 16 bits and converted to 8
> bits. (No one I know of can view a 16 bit image directly.) I then
> copied the image 5 times and in PS posterized them to 7, 6, 5, 4, and 3
> bits. I added the labels after so they're not part of the posterizing
> process. I left this as a .PSD file so you could click on each layer,
> look at the histogram, and see the number of discrete levels. What
> you'll see is that there is no improvement in the image beyond 6 bits.
> There is some degradation at 5 bits. So this TP images has 5 - 6 bits
> of real data. Beyond that it is just noise. What you're seeing here is
> 3 to 8 bits covering the full dynamic range of the image. As I've said
> on here many times, any number of bits can cover any dynamic range.
Chuck,
There's a qualifier to the above argument: film's response to light is
non-linear; if you convert your output (scanned) levels to light, your
dynamic range extends. A more faithful comparison would be between
digitised film and gamma-corrected ccd data. i think you will find that
even though the ccd has significantly better SNR, the diference is not
as large as it would seem at first sight.
radu
>
> The reason CCDs need more than 8 bits is that they're noise is much
> lower than film so in order to cover the dynamic range AND reach the
> noise floor, more than 8 bits are required. This also relates to why
> film scanner noise is not an issue except on very dense negatives
> because the scanner noise is totally masked by the film noise.
>
> In order to find out what bit depth of noise-free data you really have
> in a CCD image, you'd need to posterize the unprocessed 16 bit images
> to 15, 14, 13, etc. bits and then process the images as you normally
> would and check to see at what bit depth degradation is noticeable. I'd
> be really surprised if it 16 bits. If you can get 10 - 12 bits
> noise-free data, you really have something. I can show you weather
> satellite images that started out as 10 bit noise-free data. They look
> amazing at 8 bits even after extensive PS processing.
>
> Note: I'm only taking about viewing these images on a monitor or print,
> not scientific analysis of the raw data where you might squeeze out
> another useful bit.
>
> Chuck
>
> --------------------------------------------
>
> > >In fact, it takes a FILM person to trully appreciate what a CCD can do
> > >.... downloading 16 bit data right out of the sky is more meaningful
> > to a
> > >film person than to any CCD newbie ... what do they know about being
> > >stuck
> > >in an 8 bit world?
> >
> > Hi Tony
> >
> > I have to desagree a little with you. Film isn't limited to the 8bits.
> > Becouse it is a chemical medium, I think that the only limitation that
> > we face with the bit deph is the scanner quality. Most modern scanners
> > have a value higher than 10 bits (there are some that works with
> > 16bits), and of course they generate 16bits files. Also, we have the
> > possibility to integrate several film images to increase the SNR, and
> > in some ways the bit deph is increased too. Of course we need a
> > software that matches that potential. Before Photoshop CS we where in
> > fact limited... but now we have other softwares as well that works
> > with 16 and 32bits. I am convinced that film shots can be pushed to
> > higher levels with the new digital tools, as well with the stacking
> > procedure. There are a lot of techniques that most of film guys don't
> > use, but it is possible to apply them. Just for example think of
> > deconvolution or wavelet processing. Also we can make pseudo-flat
> > images, in a better way than lin ear gradients or bluring. Just see
> > the work that Wei-Hao is doing in that field.
> > What am I trying to say with all of this? Simple. Film shots are
> > highly subestimated, and there is a lot more to do than we have seen
> > in the past years.
> >
> > PS: Just in any case, I am not an enemy of the CCDs. In fact I have
> > used a ST7... and I follow very closely the work of guys like Daniel
> > Vershatse, Volker Wender, Robert Gendler, and of course you. I wish
> > I'd could have the equipment you use. <g>
> >
> >
> > Regards,
> >
> > Carlos Milovic F.
> > -------------------------
> > Astro & Photo - CMF
> > http://www.astrophoto.vze.com
> > -------------------------
> > Visita "AstroFoto", el foro de astrofotografía en español
> > http://espanol.groups.yahoo.com/group/astrofoto
> _______________________________________________
> Astro-Photo mailing list
> Astro-Photo@seds.org
> http://seds.org/mailman/listinfo/astro-photo
>
>
--
-------------
Radu Corlan Snail Mail: Bucuresti sect. 1,
rcorlan@pcnet.ro str. Argentina nr. 28, Romania
You can still escape the "Gates" of Hell!
Use Linux!
_______________________________________________
Astro-Photo mailing list
Astro-Photo@seds.org
http://seds.org/mailman/listinfo/astro-photo