Idaho Airships, Inc.

Mike Russell has produced an intriguing exercise in applied color theory on his Curvemeister site. I’ve reviewed Curvemeister for PAPA and submitted a strong “Recommend” at a recent lecture in Van Nuys, but haven’t yet had the opportunity to explore the latest version in depth. That will come, prior to the upcoming presentation in Redondo Beach in February.

This particular exercise addresses some issues we recently explored on PAPAChat with some color issues in a Canon 1DsMIII sample image. If you were privy to that exchange, you’ll simply have to accommodate the particulars of Mike’s tutorial being distinct from yet germane to our conclusions.

The following GamutVision (thanks again, Norman Koren!) graphic demonstrates the locus of constituent colors as they were mapped to the aRGB color space out of Mike’s camera. You can easily see that the cobalt blue Mike is referencing is located along the ventral blue ridge on this Lab plot. The loci of blue clusters positioned along the ridge represent maximally saturated colors at their respective luminance levels. Of special importance to other topics (e.g. sRGB sufficiency, PC vs. Apple Gamma, Black Point Compensation et. al.) recently discussed on PAPAChat is the density of clusters on the skin (fully saturated) in the very low key (low luminance) portion of the gamut. This will prove extremely problematic for conversion to most output profiles (click the thumbnail for a larger image):

In fact, when the aRGB image is mapped to an example output profile (in this case, a Canon i9900 dye ink printer profiled with Ilford Galleria Classic Gloss media…a very competent combination), the displacement of those saturated colors is profound. Not only are the low key blues remapped in and up (away from the extremities of the aRGB gamut), but the midrange blues are remapped almost halfway to the L axis of the color space! This demonstration is made using the Perceptual Rendering Intent with Black Point Compensation (click the thumbnail for a larger image):

Using the Absolute Colorimetric Rendering Intent, without BPC, produces a much different result-if you have GamutVision (and you should…), please view the output in Absolute and note the direction of the conversion vectors.

The remainder of the exercise is easily accomplished as demonstrated by Mike using his Curvemeister Profile Plotter. While GamutVision has the ability to plot out various problems such as Δ-E, Δ-C, and Δ-L, doing so would exceed the intent of this post and be confined to the working spaces anyway-I’m really just getting into the juicier parts of GamutVision and haven’t encountered the machinery for generating and working with specific, stand-alone image profiles.

You should now be able to digest some of the corruptive issues behind the 1DsMIII image we critiqued recently on PAPAChat.

L