From: provine@enel.ucalgary.ca (Joseph Provine)
Newsgroups: comp.graphics.algorithms
Subject: Re: Hilbert curves vs recursive blocks
Date: 2 Apr 1996 02:05:00 GMT
In article <315A1E72.4E0@softwarezone.com>,
Michael Chock wrote:
>Scott Cunningham wrote:
>>
>> Ok, I found the Hilbert curves in Gems II very interesting and jumped
>> right into playing with it. My question is this; What is this useful
>> for? I figured it was a great way to optimize graphic compression
>> (because it clumps areas together thus optimizing the similarities in
>> blocks of color better than a scanline approach) How would this be
>> better than just recursively dividing an area into quarters and reading
>> it in recursivly smaller blocks; thus:
>>
>> 1 2 5 6 17 18 21 22
>>
>> 3 4 7 8 19 20 23 24
>>
>> 9 10 13 14 25 26 29 30
>>
>> 11 12 15 16 27 28 31 32
>>
>> What am I missing in this? What advantage would it be to have this
>> connected by a continuous winding path using a more complex algorithm?
>>
>> --
>> - Scott Edward Cunningham [scottec@netcom.com] Seattle, Washington, USA -
I did some work on compressing (Peano-)Hilbert scanned data. Why go about
doing this? - it enhances the correlation between pixels (within a small
reasonable neighbourhood) as we exploit correlation in both directions -
horizontal and vertical - due to the nature of the curve. There is a preprint
at the web-site below and a colleague of mine extended this to an interesting
1-D compression scheme (using fractal geometry).
cheers,
joseph
--
e-mail: provine@enel.ucalgary.ca
url: http://www-mddsp.enel.ucalgary.ca/People/provine