Re: Cassini: where are colors and detail?

From: Chris L Peterson (clp_at_alumni.caltech.edu)
Date: 07/06/04 

Date: Tue, 06 Jul 2004 02:07:07 GMT

On Mon, 05 Jul 2004 19:22:16 -0600, "Roger N. Clark (Change username to
rnclark)" <username@qwest.net> wrote:

>Well, I guess I and many others are only doing fake science with digital
>cameras by implication of your statements. People are using them for
>spectroscopy, e.g. see:
>http://www.astrosurf.com/buil/
>http://www.astrosurf.com/buil/us/digit/spectra.htm

Sorry, it isn't my intent to in any way put down people who have found a way to
do something useful with these cameras. I can see no particularly good reason to
use a color sensor for spectroscopy, but the quality of the data is largely
independent of the quality of the filters (the result is spatial), so it doesn't
matter a lot.

>The spectral response on the 10D is well blocked over all wavelengths of the
>sesnor, and there is no leakage. See the above link. I've independently
>measured the response and agree with the above results.
>Perhaps you've had bad experience with another sensor?

My experience includes the same sensor (and others, as well). The curves are
fair, but not particularly good. The curves as shown are not useful, as the
response is shown on a linear scale. If this were plotted logarithmically, it
would be seen that all of the filters pass enough out-of-band light to be a
problem when working with deep data. Certainly, any attempt to use this sensor
for collecting images highly illuminated by narrow emission lines quickly shows
its limitations.

The details of this particular sensor are probably not entirely relevant, as it
is certainly not the type of device that would find its way onto a spacecraft.
I'm sure that any color imager would utilize a high quality CCD, and would
probably require a custom filter matrix. I'm not sure it is really possible to
produce one, however, with the necessary degree of blocking. And there is still
the problem of resolution loss. It is quite common now to make single pixel
measurements from image data; that is not possible with data utilizing a filter
matrix.

>I'm using the 10D for surface brightness photometry, e.g. see
>http://clarkvision.com/astro/surface-brightness-profiles/introduction.html

Interesting work.

>Pixel to pixel non-uniformities are a non-issue. CCDs also have
>these.

They do, but I'm talking about pixel to pixel non-uniformaties in spectral
response, which may be considerably greater (from filter matrix issues) than any
similar variation found in the underlying chip.

>Digital cameras are stable and produce absolutely superb
>pictures, even when extremely stretched. Even if there were residual
>variations, a flat field would remove them. Even CCD users do
>flat fields.

I'm not really comparing consumer digital cameras to professional CCD cameras.
I'm comparing sensors that utilize an overlaying filter matrix to those that do
not.

>> They do this in large part because of very smart internal image processing,
>> which includes a good deal of knowledge about what the image is "supposed" to
>> look like. I use a 300D myself, and I agree that it does very well, certainly
>> better than film (which is itself an even worse medium for any kind of
>> scientific work). But the Canon does not do well on things like nebulas where
>> the light doesn't look anything like a typical terrestrial source.
>
>This does not make sense. If the sensor is linear (it is), and the
>filters well behaved (they are), the system provides good data
>defined by the bandpass. Digital cameras could not "know" how
>the vast variety pictures people take would look like and process
>that accordingly. The way they take good pictures is
>basic and fundamental linear response without serious artifacts,
>just like any good scientific tool.

Actually, they do. The advances being made in digital cameras are related more
to the image processing than to the sensors. Take a look at the patents Canon
and Nikon have in these areas. They are utilizing a lot of processing to
calibrate the data. Again, this is quite evident in the fact that these cameras
cannot take good images of emission nebulas.

There is very little reason to include such a camera on any mission, since much
higher quality images and data are produced by cameras with selectable filters.
The only case I can think of where a color sensor would be useful is where the
target is moving very fast, or changing quickly with time, and individual images
through different filters could not be combined. But there have only been a very
few cases where this was an issue (close flybys). Presumably this is why no
proposals for color cameras have yet been accepted for research probes.

_________________________________________________

Chris L Peterson
Cloudbait Observatory
http://www.cloudbait.com

Relevant Pages

  • Re: Best non-SLR
    ... I can't speak for all such cameras, ... I quoted DPReview stating that they would have judged the FZ50 more ... mentioned DPReview's criticism of what the small sensor did for it. ... excellent images, just the opposite actually. ...
    (rec.photo.digital)
  • Re: Utterly off-topic: digital cameras
    ... priced cameras with good reputations. ... more gross image distortion (because the lens is smaller) ... fuzzier images (because the lens is worse and more of it ... of the sensor design rather than of the optics. ...
    (uk.religion.christian)
  • Re: New lossless image compression algorithm
    ... white images. ... really correspond to the human eye's three filters, ... The compression software does not try to find correlations between ... pairs of images (eight of the nine cameras on each rover make up four ...
    (comp.compression)
  • Re: Best non-SLR
    ... non-SLR cameras do not cover the entire gamut of such cameras, ... I have shown the images to prove my point. ... expected from a P&S's small sensor. ... if you wanted to compare the ...
    (rec.photo.digital)
  • Re: Thermal Images enhancement
    ... I do work with thermal images occasionally. ... cameras that, like the search and rescue, use live video. ... you could try edge finding with DOG filters ...
    (comp.soft-sys.matlab)