XYZ to RGB and white point
- From: "Stewart DIBBS" <sjd@xxxxxxxxx>
- Date: Tue, 12 Sep 2006 21:00:39 -0400
My spectroradiometer project generates a double precision spectrum from
360 - 830 nm from a set of monochrome images, from which I generate
tristimulus XYZ values. These values are not normalized to xyz bar values.
Converting to RGB using an appropriate matrix generates fp values which are
then scaled 0-255 for creation of an RGB24 bitmap.
Any RGB values < 0.0 are set to zero, which of course sets the black point.
The problem is deciding on the white point. If my target includes just one
saturated white pixel, I can simply use the RGB maximums as the white point,
and scale the RGB floats into 0-255. Visually it works very nicely for both
brightly and dimly illuminated targets. Put another way, if bright
Tristimulus XYZ all approx= 1400 gives a bright white, XYZ approx= 140 gives
a dim white, as expected.
A target spectrum cannot be assumed to have a saturated white pixel, and in
fact in most real cases it won't. I'm thinking perhaps I could create a
simulated white pixel based on the stats of the monochrome images (feasible
I think) or the solution is something about the illuminant I use during the
Tristimulus XYZ calculations.
Perhaps someone can offer clarification or another approach?
regards,
Stewart DIBBS
========================================
www.pixcl.com Gatineau, Quebec, CANADA
========================================
.
- Prev by Date: Re: Matched filter question
- Next by Date: Re: Matched filter question
- Previous by thread: CMU house sequence
- Next by thread: Chips for Real-Time Image Recognition
- Index(es):
Relevant Pages
|