From: Simon K. <si...@ne...> - 2007-12-31 20:00:45
|
On Sat, Dec 29, 2007 at 04:19:09PM +0100, Udi Fuchs wrote: > The first is related to Rafael test.c program from a few weeks ago. It > seems that applying the matrix in lcms gives very different results. > > Some of the difference is that lcms uses the sRGB gamma function that > has very small linear part. This adds "fog" to the picture, that I > still didn't find how to correct. I am stepping in half-understood territory (as usual) ;), but since this is a simple a simple 3*3 (or 4*3) matrix, how differently can lcms apply it? Wouldn't it be wrong for it to be messing with gamma at the same time? (I forget the ordering of all of the transforms again.) Perhaps something that would help in this area is something that visualizes (and allows realtime ending of :)) the matrix. I've been wanting to make an interface to tweaking the dcraw-based matrices for some time now, because I am not entirely happy with the dcraw ones (eg: the Canon 20D matrix has bright oranges too close to yellow, while the 10D one is great -- to the point where I am happier with ufraw output than the camera JPEGs). I've seen some other tools have sliders for "hue" and "saturation" for each of the primary colours for shadows, mid-tones, and highlights. Perhaps this is (mathematically) sufficient to describe the matrix, assuming an always-grey (neutral) axis? It should be possible to fit this in a new ufraw tab (if there is room for tab buttons...). Perhaps then matrices could be applied just as ICC profiles are -- possibly even with a stacking options for effect (eg: sepia tone, inverted channels, infrared-like, etc...) Simon- |