While I'm still waiting to be able to test this functionality in more detail, I've been checking up on the profile refinement option in the ColorMunki software.
Since several people have asked me about it I thought I'd post what I'd found out from some limited personal testing and talking with people at X-Rite (thanks).
How it works... ;-)
The first 50 patches are always the same set of device recipes (RGB or CMYK values).
With measurements from the first 50 patches, ColorMunki builds a profile and then runs a known set of Lab values (containing neutrals, skin tones, etc.) through the profile to determine roughly the device values of those desired colours.
This data is used to build the second target (which explains why the 2nd targets look quite similar)
Next, print and measure the newly created second target and then build the final profile using both sets of device recipe and measurement data.
This refinement process helps identify exactly where these key colour areas reside in the printer gamut.
In previous methods, it wasn't known which combination of device recipes actually printed neutral tones so more combinations of colours were printed/measured to find out.
The new iterative discovery process helps find these neutrals much more easily with fewer colour patches.
This is the key to the low patch count profiles but does depend on predictable behaviour of the printer in regions not sampled in the first set of patches. This predictability is both a strength and potential weakness in the ColorMunki technique - I guess we'll have to wait for more people to do detailed testing, however you can bet that the algorithms have been tested on a lot of real world printers.
The image based optimisation works the same way as the second set of 50 patches.
The CM software automatically extract colours from the image, converts them to Lab (using the source / image profile)
The new target is then printed and measured. This information is then added to the previous measurement data set, and used to build a new profile.
Using this method, more information is gained about the printer's behaviour in the colour regions that were extracted from the refinement image.
The idea is that, next time you print an image using the new profile, the results will be even more colour accurate.
Because the process uses colours extracted from the image in this process, you can improve any colour region you want by using an image that contains those colours.
Since the process just adds additional pairs of device recipes and corresponding measured Lab values into the profile data set, it is a non-destructive process and can be repeated as many times as needed without damaging the profile.
Note that like with any profile building, this is subject to accurate measurements.
Remember as with many things 'Garbage in - garbage out'. This is one reason I'd want to give refined profiles a new name, since a duff set of measurements (one blocked nozzle for example) will mess up your profile and there is no easy way to undo things.
One area I'm curious to look at, is to see how well the CM profiles handle greyscale images, although it should be noted that there are no profile refinement options (such as true greys or paper relative greys) so it is a case of you get what you get. As ever, take this in context of the target market of the device - if you know of the difference about the greys, then you are in the ColorMunki 'Super Expert' category ;-)