[FFmpeg-soc] SoC query

[Mike Melanson]

Deep Pal wrote:
> Of course I understand that, say one applicant mentions now and the other
> mentions later and both qualify through the qualifying task. Will the
> guy who mentioned
> earlier be getting more priority.

The formal process works like this:

* All interested students submit entries through Google's web interface
starting on March 24.
* The FFmpeg mentors privately rank and comment upon the various
applications in Google's web interface.

> Yes, which raw files to start encoding with.

Any movies you wish. I know of raw YUV reference streams, but they are
interlaced. And they're huge to download.

I'm not certain about this, but perhaps you may wish to start with video
compressed with VQ methods since the data hasn't been quantized in the
conventional JPEG/MPEG way already. Just a feeling; someone will call me
on this if I'm wrong. Here are some VQ samples:

  http://samples.mplayerhq.hu/game-formats/idroq/
  http://samples.mplayerhq.hu/V-codecs/SVQ1/

> And also how to go about with the testing?

My first impulse would be something like this:

* Transcode video to sequence of JPEG files as a baseline reference:
- mkdir baseline-reference:
- ffmpeg -i <video> baseline-reference/%05d.jpg
* make your changes, recompile
* Transcode video to a new sequence of JPEG files:
- ffmpeg -i <video> optimal/%05d.jpg

Are the files in the optimal/ subdirectory valid (use a viewer to
visually inspect)? Are they consistently smaller than the files in
baseline-reference/? If so, then you might have achieved the goal.

And when you get to this point, we should probably run another test to
ensure the various files are equal after entropy coding (I think they
should be).

	Hope this helps...
-- 
	-Mike Melanson