[Tig] Is there anybody out there

[TSassoon at aol.com]

In a message dated 8/3/06 12:01:05 PM, bfriesen at simple.dallas.tx.us writes:

> Often the number of bits used to support digitized values is (mistakenly) 
> equated to "dynamic range"
> but the number of bits only has to do with precision and not the range.
> 

Bits can be used for either or both precision or dynamic range. Consider 
Cineon - is there something magical about putting the black point at 95, or 90% 
white at 685? Definitely not, it's just a choice to give a generous amount of 
room to recover from possible poor DMin calibration. Quiz - how many bits 
approximately are used to describe the straight-line portion of the response curve 
in a 10-bit Cineon file?

The more noise there is in a system, the less bit-depth one strictly speaking 
needs. Dithering lacks compression efficiency, but conserves bit-depth. If 
half the pixels in an area are 127 and half 128 (8-bit), randomly ordered, one 
perceives a 9-bit intermediate value. In a 75/25 split, it's now a 10-bit 
value, and so on. I mean storage values of course, not calculation environment.

Quiz 2: what's the minimum percent noise one adds to turn an 8-bit image into 
a 16-bit image (in a 16-bit/float environment)?


Tim Sassoon
SFD vfx & creative post
Santa Monica, CA