Le 31 juil. 08 à 22:38, dda a écrit :
Basically, de-interlacing is moving lines of pixels around, right?
So all it boils down to is memory copy operations. I am sure I have
some assembly code for that somewhere... ;-)
It's not that simple. What I'm trying to do here is to deinterlace
video pictures and yet kept the more details I can. These Pictures
have been shot in DigitalBetacam (Currently the topmost recording
format in Standard Definition video). Then digitized to disk in an
uncompressed format. And will then be compressed in various format
depending on your clients request.
The simplest/fatest way to deinterlace is to remove the odd/even lines
of pixels and replace them by the upper or lower remaining lines.
For pictures that does not include any graphics or special effects,
one's eyes can't barely see the difference if they aren't trained to
picture quality control. But when it comes to titles and GFX (Or even
charts !), the result is just not acceptable. That's why I need to
interpolate the details I removed, and keep the image uncompressed all
along the process.
But for a 50 minutes movie, wich is more or less the standard duration
for us, we will have 87 GB for the original, 87 GB for the Spool and
87 GB for the output. It's just 'Gargantuesque' ! (Hugely huge !!!)
But as I'm writing these lines, I tought of a new solution. Processed
the pictures in small chunk of data. When the chunk is processed,
insert it in the output movie. Commit to the output file. Delete the
chunk. start to process a new chunk. It sounds more efficient to me.
And if ever the machine crashed, I will just have to restart from the
last saved position. I'll try it tomorrow.
Good night or good evening to all of you.
Eric.
_______________________________________________
Unsubscribe or switch delivery mode:
<http://www.realsoftware.com/support/listmanager/>
Search the archives:
<http://support.realsoftware.com/listarchives/lists.html>
|