View Single Post
Old July 15th, 2011, 07:13 AM
Doug Kerr Doug Kerr is offline
Senior Member
Join Date: May 2006
Location: Alamogordo, New Mexico, USA
Posts: 8,611

Hi, Bart,

Originally Posted by Bart_van_der_Wolf View Post
One could also argue that a colorspace doesn't have a dynamic range, but that the encoding precision allows to capture a certain range.
Oh, quite so. I always try to maintain that distinction in my writing (but in some cases I don't, just for the sake of "colloquy").

As a small side note, here is some information about other HDR file formats, their Dynamic range capabilities, and their precision.[/quote]

Thanks for the link to that most useful paper. I think I have not seen it before (although the "spiral slice" illustrations - Fig. 15 in particular - seem strangely familiar - perhaps the format is not original with that author).

It also stipulates that while the dynamic range capabilities are more than adequate for most purposes, the precision of a 96 bits per pixel floating point TIFF is a bit of a waste because probably most of lower significant bits will just hold an accurate encoding of noise (given the input sources).

[quote] It also makes for poor compression capability. That's probably also why the standalone Unified Colors software allows to save a (lossy ;-)) compressed version of the images in the BEF format. /QUOTE]
Yes, and I'm not sure what kind of non-reversible [;-)] compression they use.

Circling back to the notation front, I note that Holzer uses "dynamic range" to mean the encoding range of a coding scheme, and uses "accuracy" when "precision" is probably meant.

In the Bef scheme, it is not clear that the 32-bit floating point representation is really warranted for the two chromaticity coordinates e and f (that plane is in fact chromaticity, not chrominance, as I understand the derivation of the coordinates). It seems that Bef can encode about 4.6E+18 different values of chromaticity.

Thanks again.

Best regards,

Reply With Quote