Again, you need to consider what you want to run through the profile to compare. You may want to use the iStar target. See: http://www.aardenburg-imaging.com/
Mark hangs out here, he might see this and comment on it’s further usage. If you are using color patches to evaluate that are known to be outside the gamut of the profile, well that’s kind of a waste. So the patches you select can play a role in the dE report you get from ColorThink.
There's an early version of the iStar target that Andrew refers to on the Wilhelm Imaging Research website. You can find it here:
http://wilhelm-research.com/istar/index.htmlThis html page has numerous articles and even some beta software that can calculate I* color and tone values. (note: this software is not for the faint of heart! It was originally developed with color scientists in mind and end-user licensing fees that would pay for advanced training sessions).
I made a newer version of the the target that adds some ordered skin tones and the 24 patch Macbeth color chart to the target page. You can find this newer target at:
http://www.aardenburg-imaging.com/acceleratedagingtests.htmlYou will need to register on the Aardenburg website (it's free). Then you can download the AaI_testing.zip folder. The color target in this folder is the newer version of the iStar target.
Like Bill Atkinson's visually appealing profiling targets, the iStar target presents a visually ordered set of colors that the human brain can more easily interpret subjectively without resorting to numeric analysis. At the other extreme are scrambled color patch targets that make absolutely no sense to the human observer! In particular the iStar target presents 12 individual hue planes with graduated colors going from dark to light and from low chroma to highest chroma that still fits in gamut within the sRGB colorspace. That is the basic design concept behind the target. Note also that there's a blue color quadrant that shows hues deviating a little bit purple while some are a little more cyanish-blue in color. This is because the LAB colorspace, although very very good in terms of perceptual uniformity, does have some visual non uniformity particularly in the blue hue range. Still, it's the best color model for the human visual system we have at this point in time. This blue-purple deviation within the constant LAB hue plane for blue colors is commonly referred to as the "blue turns purple problem" when trying to use ICC profiles to print out of gamut blues to within gamut. Various profile software packages attempt to correct for this error and other tonal issues in the perceptual tag (by brute force remapping of those LAB colors to other LAB values that look closer to what humans interpret as a constant blue hue over varying chroma and lightness levels), Nonetheless, the iStar target presents this blue LAB hue plane to the output device to evaluate how it renders.
The I* (iStar) metric is too complex a subject to present here in extensive detail, but I will try to explain some basics, anyway. The concepts of this I* metric are really important for printmakers and for any application where color and tone reproduction is being evaluated. The I* metric has both a contrast function and a chroma weighting function. Both functions are hugely important when evaluating photographic tone reproduction. Delta E does not correctly weight chroma within the context of a complex scene (i.e., humans prioritize low chroma colors to identify scene color balance) nor can Delta E rate global or local changes in image contrast whereas the I* metric does both. The I* metric also evaluates overall tone and color reproduction in a meaningful way even when there are large color matching errors involved, i.e, which is almost always the case when trying to reproduce a larger dynamic range image onto a reflection print medium with much smaller printable dynamic range. Here's a classic example: You have an RGB 0,0,0 black value in your image file that is tagged with a working space profile as L* = 0, but you are printing to a matt fine art paper that can only make a black no darker than L* = 15. Even if you reproduce the a* and b* = 0 values perfectly on paper, you are still left with a delta L* discrepancy of 15 under best case scenario. Thus, the delta E error will also be 15, ie., a huge color matching error. What this all means for photographic prints is that average delta E errors will be huge in many circumstances when comparing image file to print (or even monitor color to print). So, how do we perceive a satisfactory color and tone reproduction on any reflection print? That's where the I* metric comes into it's own. It evaluates proportional color and tonal response which is also what humans do when looking at reflection prints. Optimum tone reproduction occurs by keeping relationships between various colors and tones in the image relatively proportional when they can not be exact. Thus, the I* metric can successfully evaluate large color and tonal changes on an easy-to-understand percentage scale whereas delta E just generates huge error values when exact color matching can't be reasonably accomplished.
The concepts I just discussed may be better grasped if you also decide to read the following paper (and other articles) on the Aardenburg Imaging website:
"Case Study #1 Evaluating the Influence of Media on Inkjet Tone And Color Reproduction With the I* Metric" (
http://www.aardenburg-imaging.com/cgi-bin/mrk/_4625ZGxkLzBeMTAwMDAwMDAwMTIzNDU2Nzg5LyoyMA==)
Although the iStar target that Digitaldog kindly mentioned in this thread is indeed a very useful generic test target designed to exercise the I* metric's color and tone algorithms effectively, the I* metric can also be used to evaluate color and tone reproduction in essentially any image. The paper just cited shows a case where a specific pictorial image was subsampled to make a custom color target which could then be read on a standard spectrophotometer. The image and the custom target were then taken through a tone reproduction exercise on three different paper types, one glossy, one mat,and one plain paper using the same printer and inks. I hope you find the I* scores for the different papers interesting.
What remains to happen for the I* metric is more interest to be shown by the color and imaging science community and the printing industry. With such interest we could produce some software apps that will make it easy for a wider audience to use the I* metric on a practical basis. Modesty aside, the I* metric is a powerful analytical method for color and tone reproduction accuracy problems, but much to my surprise, it has been essentially ignored by the color and imaging science community and by the printing industry even though it's open source and was published more than five years ago in the NIP20 conference proceedings of IS&T. I would think the printing industry in particular would welcome a much better algorithm with a familiar percentile scale to better communicate color and tonal accuracy issues with customers. They have customers who often wonder why their digital files don't print exactly like they look on a monitor and then blame the print shop.
Time will tell if the I* metric eventually finds an interested audience.
Other technical articles on the I* Metric are also available on the documents page of the Aai&A website:
http://www.aardenburg-imaging.com/documents.htmlSorry for the long-winded post, but you can blame Digitaldog for mentioning iStar
kind regards,
Mark
