On Wed, Dec 23, 2009 at 4:30 AM, Graeme Gill <graeme@xxxxxxxxxxxxx> wrote: > Pascal de Bruijn wrote: >> >> color cast. Since Firefox 3.5 doesn't use LCMS anymore, I'm guessing >> this isn't a CMM issue. > > Possibly. Other explanations might be: > > The white point is too different from the black point, and you > are noticing the transition from the neutral target to the black > point (ie. with the LCD default -k 0). Shifting the white target > to a hue nearer that of black may help, or using a -k > 0 > may help at some cost to the contrast range. > > It might be an optical effect (although your screen captures > hint otherwise). I've noticed some displays have quite > a hue shift with viewing angle near black, particularly > if the color is not naturally neutral. A larger -A parameter > may help in this situation (assuming -k 0). > > It may be that the target curve shape is too much of a stretch > for the display near the dark end. This may be because the > current default is to use an input offset model of gamma curve > shape. You could try all output offset as an alternative (-f 1). > [I'm thinking of making this the default in 1.1.0 RC3]. Well, -f 1 seems to reduce the problem quite a bit... But it's not completely gone... The -A parameter did not really seem to make a difference at all. Maybe I should try to create an XYZ LUT profile. Regards, Pascal de Bruijn