2005.05.25 13:31 "[Tiff] Re: Another libtiff+libjpeg problem", by Jean-Yves Le Ridant
KW> if your after a slightly better approximation but not too slow when KW> doing bit depth promotion you should fill the bottom bits with the
KW> upper bits...
KW> e.g. 12 bits to 16...
KW> 12 bits names
BA9876543210 ->> BA9876543210BA98
KW> The above being a bit shift
KW> out := (in << 4) | (in >> 8)
KW> 8 bits
76543210 ->> 7654321076543210
KW> it works and ensures that peak input is mapped to peak output.
Yes but...., unsigned shift is perfect for a and b of ICCLAB, and signed shift would be right for those of CIELAB ( and guys have strange ideas as jpeg compressing of CIELAB ...) :-)
No hope, really handling 16 bits is a job, and would be better handled by jpeglib interface, mixed with/in existing tables. So at this point it's obstensibely a quickwatch.