[Zlib-devel] Re: [png-implement] Libpng-1.2.2rc1 and libpng-1.0.13rc1

[Greg Roelofs]

Mark wrote:

> On Wednesday, April 10, 2002, at 08:54  AM, Greg Roelofs wrote:
> > It appears that memLevels 7-9 have an identical effect on Huffman-only
> > compression,

> That would be expected for small files.  The memLevel variation for 
> Huffman-only coding can only manifest if the file is long enough or the 
> blocks short  enough for the file to be broken up into more than one 
> block.  In this case, that doesn't happen until the blocks are 4K or 
> less.

OK, that makes sense.  This is a 256x8x3-byte (6K) stream.

> > However, I did see at least one case where an intermediate level was
> > best:
> >
> >    2122340:  5  2  3,0,0,2,1,2,3,0   5754  <--

> This is telling me that that particular example is changing the 
> distribution of its bytes approximately every 2K bytes.

It's a section of a truecolor photograph (scanned directly to PNG), and
each row is 768 bytes wide, so every 2K would 2 2/3 rows.  Is that close
enough?  I would have expected something to happen on even row boundaries.

Anyway, I can redo the experiment in reasonable time by using a wider
section of image.  Taller ones aren't really practical...

Greg