Re: Image Delta Compression

On Wed, 3 Dec 2003 03:50:16 +0100, "Marton Anka" <marton at 03am dot
com> wrote:
>A good analogy would be screenshots: let's say you're writing an email and
>you take screenshot, then take another one in 30 seconds. The difference
>would usually be very small - only a few lines added, and no parts of the
>image would move. However, you could also scroll the contents of the email
>by two lines up between the two screenshots, in which case you'd have a fair
>amount of movement to handle. Or you could drag the window you're writing
>the email in 5 pixels up and 10 to the left - again, motion to detect.
>
>As I wrote, my initial approach - the generic binary diff - produces a
>pretty small delta output when working with small string lengths, but the
>time it takes to do so is way too long.

While they don't recognize scrolled areas (AFAIK), the algorithms in
VNC do a good job of minimizing updates for real-time transmission of
screen changes. VNC source code may be a good place to look for ideas:

http://www.realvnc.com

If it's necessary for what you're doing, recognizing horizontally or
vertically moving areas would not be all that difficult (though it
would require a lot of additional CPU time). Detecting objects moving
in arbitrary directions would be particularly expensive.

- Sev