🚀 go-pugleaf

In article <1992Apr16.221510.7499@csi.uottawa.ca> cbbrowne@csi.uottawa.ca (Christopher Browne) writes:
>The idea of having 16 million colours (24 bits worth), that are simply
>"chosen" rather than using a palette sounds neat, in SOME ways, but is
>it ultimately ALL that great?
>
...
>
>1) It means that if some of those colours never get used, whether in a
>particular application, or possibly EVER (who'd want THAT putrid shade
>of green?), then that part of the implied palette is wasted.  It might
>be a better use of resources to have a smaller palette that contains
>the colours that are actually USED.

The main advantage of a "palette-less" system is in image processing.
With a direct RGB (or other component) method of image storage, one may
process the image by standard, linear filtering methods which use
superposition of image intensity and color information.  Even non-linear
techniques are more straightforward.  With a palette system, much translation
and untranslation is required, muddling the algorithms.  I would venture
that with a palette system, most (image processing) programmers would keep
an essentially RGB version of the image in memory for processing and
convert it to the palette-ized version for display only.  In this case
the palette is inconvenient and doesn't save any memory.

This is analogous to 12-bit logarithmic encoding of audio as opposed to
16-bit linear.  In many ways, 12-bit system can have comparable fidelity
and saves memory/storage.  However, the linear system has many advantages
in filtering and other signal processing.

-- Scott Willingham // willing@icsl.ucla.edu