[argyllcms] Re: Best way to proceed?
- From: "Mike Russell" <groups@xxxxxxxxxxxxxxxx>
- To: <argyllcms@xxxxxxxxxxxxx>
- Date: Sun, 1 Jun 2008 21:37:01 -0700
From: "Ben Goren" <ben@xxxxxxxxxxxxxxxx>
[re matching wood colors using colorchecker as reference]
I don't think that'll help me. When I've adjusted curves, etc., in
Photoshop to get the ColorChecker close to spot-on, the wood hasn't
gotten any closer -- it's just shifted to other equally-worng colors.
I can think of a couple of reasons for the problem you are seeing.
The biggest problem that I see is that you are matching colors that are of
no importance in your image - pure blue, yellow, green, cyan, etc. using
curves to match colors that have no relevance to the subject matter of your
image. For example, adjusting the bright yellow patch of the checker may
cause bends in the red, green, and blue curves extend to other, darker parts
of the image - this is probably the biggest effect.
Another problem is your choice of color space in which to curve. You are
presumably using curves in RGB space, so each curve will change the
brightness as well as the hue, and the effect of evan a small change near
the end of a curve can have large effects on the midpoint of the curve. If
you are seeing wild bends in any of the curves, then adjust only hue and
saturation, and ignore the luminance information.
You mentioned the irridescence of the wood in a later post. Fabrics have a
similar problem, being sensitive to both direction of view and lighting.
The solution used in this case is to create a packed bundle of strands, and
cut them off square at the end, and measure the end. This provides a
consistent diffuse measurement.
The fabric procedure suggests that measuring the end grain might provide a
similar stable hue measurement, that would reflect the correct overall color
of the wood. Including a shot of the end grain, shot under the same
lighting conditions as the main product shot, and matching that hue and
saturation in Lab space to your colorimeter mesaurement, by adjusting the a
and b curves. Then use the Lightness curve to add contrast and bring out
the artwork, perhaps by matching the Lightness (ignoring the a and b values)
of several representative points.
I noticed also that your background has a magenta cast - this is undesirable
because it will give the wood a more neutral appearance in your image, and
undercut your attempts to match the colors. the brighter background is best
adjusted in RGB space.
I've also tried using the i1 Scan Target to create a profile, again with
no luck -- see an earlier thread. (I've played around with it more since
then. Even with an evenly-lit square-on shot that produced a profile with
reasonable numbers, it still didn't do anything for the wood.)
I think a target consisting of neutrals, and a representative set of wood
color patches would give you a more satisfactory result than a general
purpose color target. The reason for this is that you want the deltaE of
the wood hues to be as small as possible. The accuracy of other colors,
particularly saturated ones, is much less important, and will compromise the
wood colors.
Mike Russell - www.curvemeister.com
Other related posts:
