[mythtv-users] Transcoding, mpeg-2, mpeg-4 ASP vs. AVC

[Stephen Boddy]

I'm getting a little bit confused as to what MythTV is doing on the 
transcoding front.

I'm recording UK SD DVB-T mpeg2 streams. I'm struggling to keep up with the 
stuff I'm recording, so I'd like to transcode where possible.

I've got the system transcoding files, I know it is using the "From MPEG2" 
profile. (When I change the bitrate the transcoded filesize changes.)

I understood mpeg-4 ASP to require approx. half the bitrate / filesize that 
mpeg-2 does for an equivalent quality file. Then mpeg-4 AVC is even better, 
requiring perhaps a third what mpeg-2 does.

I tried adjusting the settings, but I'm seeing a significant drop in the 
quality, even with a modest reduction in bitrate.

From this I have a few questions for anyone willing to give it a shot.

1. Should I be seeing a drop in quality when I lower the bitrate? To 
illustrate: A 1.5 GB mpeg-2 recording, transcoded to mpeg-4 with a 2200 
kilobits/sec has roughly the same filesize and quality as the original 
mpeg-2. Lowering that to 1600 kbps with high quality and 4mv enc. reduces the 
file size (prob. to about 70%) but the quality suffers quite badly. 
Background areas (like trees) that are fairly clear in the original become 
smeared blurry blobs that pan in jerky little steps. Movements of foreground 
objects (i.e. a shoulder and head shot) result in a blocky pixelating effect 
as the face moves quickly, but then settles down once the movement stops.

2. So why am I not able to get better compression for the same quality when I 
transcode to mpeg-4?

3. Does the transcode process actually use the better AVC codec? I took a look 
at the details on the ffmpeg site (home of libavcodec) and according to: 
http://ffmpeg.sourceforge.net/ffmpeg-doc.html#SEC19 they don't appear to 
support H.264 which is essentially what AVC is.

4. If not, what is the high-quality encoding for? Is it a two-pass vs. 
one-pass?

Thanks in advance for any explanations.
-- 
Steve Boddy