On 2006 May 26, at 2:09 PM, Gerhard Fuernkranz wrote:
> Ben Goren wrote: > >> A low -r value (but, of course, not so low as to have real >> trouble with noise) may produce more accurate colors at the >> expense of smoothness in photographic images. A high -r value >> will result in colors that're less accurate, but perhaps not so >> inaccurate as to be noticeable outside of a side-to-side >> comparison. I'm pretty sure from some preliminary testing, >> though, that too high an -r value will start to cost you fine >> detail, so it may remain something that needs to be tweaked >> individually. > > Ben, I'm objecting the statement "a low -r value will produce > more accurate colors". That's basically not the case. A low > smoothness factor will indeed result in a model which fits the > _training set_ with a low error.
Well, that's why I used ``may'' instead of ``will,'' and ``not so low as to have real trouble with noise.
What I'm driving at is that, within the range of values for -r which are at least large enough to be an accurate reflection of the error bands of the data, the lowest value should, I think, if I'm understanding it right, produce the most accurate colors. It's also the value most likely to bump into non-linear properties of the printer. At the other extreme, a value that's ridiculously high will produce results more akin to not even profiling the printer in the first place.