Re: Image Delta Compression

In comp.compression Marton Anka <marton at 03am dot com> wrote:
> Yes, that's what I'm thinking about doing - but I'm wondering if there's a
> fast, proven
> algorithm for aligning the blocks. The method I've described in my initial
> post sounds like a brute force approach to me. I'm not that familiar with
> MPEG itself.

  Well, one obvious answer is to compute correlation coefficients using a 
(two-dimensional) FFT.  If blocks tend to move without any corruption, 
another obvious solution is to suffix sort the two images, find (for each 
pixel in the new image) the position in the old image from which the longest 
string of matching pixels is found, and try to reach a "consensus" amoung 
the pixels within a block.

> By the way, I wasn't able to find the discussion you mentioned regarding
> X-Windows.

  It wasn't here... it was a paper I read about five months ago on the web 
somewhere.  But I can't find it either at the moment, so I don't think my 
vague memory is very helpful here.

Colin Percival