The various output image formats that are currently supported all require
POV-Ray to throw away information; not only in reducing color channel
components from floats to 16- or 8-bit integers, but also in clamping the
components to the [0, 1] range.

MegaPOV supports an RGBE/.hdr output format, and while that avoids the
clamping, it still entails a loss of precision so that each pixel can be
stuffed into four bytes.

So, given that POV-Ray already bundles libtiff and supports TIFF image maps,
I am wondering why TIFF is not supported as an output format---specifically
in its floating-point form, as that covers ground that the other formats
can't handle, and so would not be redundant.

This would, of course, be a boon for HDR rendering, and I could imagine it
being useful in scientific/research contexts. It matches POV-Ray's internal
pixel format (array of COLOUR -> COLC[5] -> float[5], er, maybe toss one
channel), so you don't lose any information when using it.

I'd like to ask if there are any points against this, and if not, then if a
patch to enable fp-TIFF output would be of interest. Much of it would track
the HDR patch for MegaPOV (e.g. disabling color clipping when this format is
chosen), and most of the new code would be in source/tiff_pov.cpp.

Post a reply to this message