Re: Which scanners REALLY provide 36 bit output? HP?

From: Steve Underwood (
Date: Tue Dec 12 2000 - 17:31:30 PST

  • Next message: Tom Martone: "Re: testing BH on copiscan 8080d"

    Stephen Williams wrote:

    > said:
    > > (Yes, it's 3 - the luminance sensors (the 'rods') have the same
    > > sensitivity as the 'green' channel 'cones')
    > Not quite, they respond to a wider spectrum then the green sensors.
    > Though they are highly sensitive to green, they respond to most all
    > of the visible color range. I think.

    > (Actually, after saying that I'm not really sure how wide the spectral
    > response is for luminance, but I do know that luminance calculations
    > certainly all include some red and blue, and I presume there is a
    > psychological/physiological reason for that.)

    Its not just about spectrum, but about resolution and sensitivity.

    The luminence sensors are much denser than the chroma sensors. They are
    most sensitive at green, drop off badly below red, and drop off much worse
    in the blue.

    The green/red sensing is at a much lower density than the luminence, but
    has fairly good sensitivity. NTSC and PAL colour TV take advantage of this,
    and use only about 1/4 of the bandwidth for the chroma info, as they use
    for the luminence info. The density and sensitivity of the blue sensors are
    very very poor. A pure red or green TV image looks bright and clear. A pure
    blue TV image looks dark and noisy. That is mostly due to the limitations
    of the eye, although the blue emitter in a TV tube is also poorer than the
    red or green..

    The bottom line is we cannot see colour in detail. Try grouping a lot of
    tiny varously coloured images together. You can still the fine detail of
    their shapes, but they all appear white.

    When you try to scan a colour photo or magazine print you are using three
    sensors which don't accurately match the spectral response of the eye, but
    have peak responses in roughly the same places. You are scanning images
    made of three dyes which strange spectral responses that only *very*
    roughly approxiamate the response of the eye's sensors. You will display
    the result on a CRT, LCD, or printer, with their own set of very rough
    approxiamations to the right spectral response. In addition, some of these
    steps are working in cyan, magenta, and yellow, rather than red, green and
    blue. There are further limitations in colour accuracy caused by this
    swapping of primary/secondary colour mode.

    If the eye's colour sensing worked in an absolute way the result would
    probably always look awful. However, the eye senses colour only in a
    relative way. Basically, an averaging process takes place across the entire
    field of vision, and the eye assumes this average to be white. All colours
    are rendered within the brain's signal processing according to this
    averaging. Now, if the sun changes to flourescent light as night comes we
    don't see the world turn the awful blue which most flourescent tubes
    produce. We see just a minor tinting of the colouration of the scene. More
    puzzling, and not fully explained, is why we appear to see a somewhat
    colourful image under low pressure sodium lamps. These produce extremely
    narrow band monochromatic light. We see zero colour variation in the
    relected light from any part of the scene, and yet we don't see the world
    in pure orange. The eye is clearly cooking up some fake, but realistic,
    colourfulness. How that realism works is a mystery, as we seem to detect
    the approxiamately correct colour of objects we have never seen in white

    Unless you understand these things, imaging will always be disappointing.
    Without a feel for the limitations of the chain of factors resulting in a
    final image, the nature of that result tends to be an unpleasant suprise.
    Often, the only way to prove to people just how bad something will look is
    to produce a sample! A sad, but necessary, waste of effort.


    Source code, list archive, and docs:
    To unsubscribe: echo unsubscribe sane-devel | mail

    This archive was generated by hypermail 2b29 : Tue Dec 12 2000 - 18:52:31 PST