Re: [dev] Looking for simple, alpha supporting image format

From: FRIGN <dev_AT_frign.de>
Date: Fri, 18 Jul 2014 20:57:34 +0200

On Fri, 18 Jul 2014 13:45:50 -0500
Samuel Holland <samuel_AT_sholland.net> wrote:

> Then put it after the magic string. And they don't have to be
> characters; they can be one-byte integers, which would be better
> anyway. My point is that the format shouldn't be limited to just 8888,
> especially if you're concerned with file size, so the header should
> specify the bits per channel somewhere.

I can see your motivation, but this is way too complex.
If you have no alpha-channel, the compression will take care of that,
because the pattern is really simple (every fourth is a 0, xxx0).
Mixed length for each channel just doesn't make any sense to me (or is
at least confined to special applications).

Here also applies the argument I gave on the alpha-channel: It may not
work on mixed precision, but if your image for instance doesn't have a
red and green channel, the pattern is (00x0). And yet again, the
compression will easily handle this without us having to pay with any
loss in simplicity of the format.

Using variable length-patterns looks to me more like a damn block-cipher!

Cheers

FRIGN

-- 
FRIGN <dev_AT_frign.de>
Received on Fri Jul 18 2014 - 20:57:34 CEST

This archive was generated by hypermail 2.3.0 : Fri Jul 18 2014 - 21:00:10 CEST