other tests: first I must apologize, Cedric has a point.
Comparisons are way more complicated
My initial tests (conversion to CMYK, output jpg and tif: comparison between the 2 files)
are basically confirmed, so the issue remains more or less.
But, my quick conclusion that creating a jpg and a tif from the same image, outputs almost identical files (one jpg and one tiff) is not precisely correct
It seems, and it is also logical, that it depends on
if the input image is jpg or tif (tif in input tends to produce more difference)
the compare app: textcompare which I trusted, most probably is not really reliable
quality of the image: how much, difficult to say, but must have some impact
And this probably makes the issue of a big difference in CMYK conversion , somewhat less odd etc.
Strange, since the output comes from same image, but if we’d assume that CMYK might be very picky and detailed etc. then something similar to having a tif in input (as per above), could happen.
Still, a difference close to 100% seems very high
Sorry Cedric, you were basically right. Even if it depends on many parameters
Hmm, they deleted their account, so may not read this, but here goes anyway.
The reason for the differences in the color values between the sRGB and CMYK files is actually very simple and has nothing to do with file types (assuming the comparison is done correctly).
Here is a good explanation, that also shows the shapes of the sRGB and CMYK gamuts:
So, to summarise: Unless the colors in the original sRGB file have been carefully chosen, a significant number of pixels will be changed in order to fit within the smaller CMYK gamut. There are different ways to handle the conversion, which all tend to give slightly different looks. A good program will let you choose which intent to use, but none will give you a 100% faithful sRGB → CMYK conversion, since that’s simply impossible in most cases.
Here is a 3D depiction of the CMYK gamut (solid shape) inside the sRGB gamut (wireframe):
One thing that struck me now, but hasn’t before when I’ve seen this illustration of the various color gamuts, is the following:
The RGB gamuts are triangles, and I’ve always been thinking that this follows from them being defined by the three base colors. But CMYK, which is illustrated with five corners, has at the outset four “colors”. The black element may possibly be seen to work in a third dimension, so that for this 2D illustration CMYK may have no more than three defining “poles”. But in any case, (is there a short answer as to) why five corners for the CMYK gamut?
In the on-line app linked to, the gamuts are drawn in CIELAB space. In that space, RGB gamuts are not “triangles”. For example, sRGB has six “corners” in the 2D view.
