Re: Q: Which 2 wavelengths give unique colour in CIE-Yxy space?
From: Tom H. (tom_reader_at_att.nospam*.net)
Date: 02/16/05
- Next message: William Snyder: "Re: Extract skeleton"
- Previous message: Dimitrios Iakovou: "Re: histogram analysis"
- In reply to: Dave Martindale: "Re: Q: Which 2 wavelengths give unique colour in CIE-Yxy space?"
- Messages sorted by: [ date ] [ thread ]
Date: Wed, 16 Feb 2005 00:04:54 GMT
I agree with Dave that to do the job accurately requires knowledge of the
reflectance spectra of the objects in the image, but there are probably ways
to approximate color temperature change. Digicams have adjustments to
compensate for different illuminants, so there should be a way to back that
compensation out and then apply one of your own making. I suggest doing a
google on color temperature and digital imaging or cameras and you might be
able to find out what others have done to do something similar to what you
want to do. For example, there may be software for adjusting digital images
that is in effect changing color temperature, but is just not putting it in
those terms. Photoshop is too expensive for my likes, so I don't know what
it can and can't do, but it may even have a color temperature adjustment
tool. Then we have to get into the color gamut of the imaging device (i.e.
computer monitor or printer --there comes the CMYK)......I'm not going
there.
Tom
"Dave Martindale" <davem@cs.ubc.ca> wrote in message
news:cutnet$faj$1@mughi.cs.ubc.ca...
> "Martijn" <m.w.hendrikx@student.utwente.nl> writes:
>
>>Is it possible to relate a unique position in the CIE-Yxy-chart to 2 or 3
>>wavelengths (given on a nanometer scale) that yield this specific colour
>>when combined? I am looking for an algorithms to switch from CMYK or
>>CIEYxy
>>space to wavelengths because I would like to perform calculations on the
>>physical spectrum of an image. So, I am now thinking of attributing two or
>>three wavelengths to every pixel in my image, finding the appropriate
>>levels
>>of intensity of each wavelength so that combination of these two yields
>>the
>>original colour. This way, I can represent the colours in the image on a
>>intensity versus wavelength graph.
>
> Well, determine the CIE XY coordinates for your colour. Then draw a
> straight line in almost any direction that goes through the point (X,
> Y). Note where this line intersects the horseshoe-shaped line of
> monochromatic colours. If it intersects the monochromatic colour line
> in two places, then a combination of those two colours in the right
> ratio will yield your original desired colour. Each different slope of
> line will give you another pair of monochromatic colours. (But some
> near-vertical lines intersect the monochromatic colour line at only one
> point; if that happens you can't use that line).
>
> However, this approach isn't that useful in general. Things that are
> more likely to be useful:
>
> 1: Pick a white point, and find its (X,Y) point on the CIE diagram.
> Then, for an arbitrary colour, draw a straight line from the white point
> through the (X,Y) coordinates of your colour, and extend it to where it
> crosses the monochromatic colour line. The monochromatic colour at that
> crossing is the "dominant wavelength" for your colour. It's a way of
> associating a unique wavelength with each possible colour (except
> purples, see below). It also means that you can recreate your original
> colour by adding single-frequency light and white light, both of which
> have rather simple spectra.
>
> (Purples don't work because the line from white doesn't intersect the
> monochromatic colour locus. But you can extend the line in the *other*
> direction to intersect the locus, and determine the wavelength of the
> complementary colour).
>
> 2: Pick three pure monochromatic colours spaced around the CIE diagram.
> With them, you can reproduce any colour in the interior of the triangle
> defined by the locations of the 3 colours on the CIE diagram.
> A red, a green, and a blue will give you the largest area, and thus the
> most colours reproduceable, but any 3 colours not in a straight line
> will work for some range of colour.
>
>>To be a bit more specific about the objective: I want to make a program
>>that
>>induces a shift in colour-temperature. Since the colour-temperature can be
>>adjusted to 'move' peaks in the wavelength spectrum to the left of to the
>>right, I prefer to convert my image data from CMYK/CIEYxy to wavelength
>>before changing the image.
>
> To *really* model the effect of a change in colour temperature of
> illuminating light, you need the full reflectance spectra of the
> objects, or the full spectra of the light reaching the camera. A RGB
> image has thrown away most of this information. The best you can do is
> just adjust the relative amplitude of RGB. So I don't see how you're
> going to model the effect of a colour temperature change.
>
> Dave
>
- Next message: William Snyder: "Re: Extract skeleton"
- Previous message: Dimitrios Iakovou: "Re: histogram analysis"
- In reply to: Dave Martindale: "Re: Q: Which 2 wavelengths give unique colour in CIE-Yxy space?"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
|
