All things related to multicast

["Rich Griffin (CampusEAI Consortium)" <>]

Frank,

What H.264 encoder are you using? Different encoders expose various 
parameters - some of which may help in low-bitrate encoding. 
Additionally, for low bitrate encodes, there should be some care in 
pre-processing if that is an option. Avisynth (www.avisynth.org) can 
help a lot with judicious use of various filtering, de-interlacing, etc. 
prior to encoding in, say, x264 
(http://www.videolan.org/developers/x264.html).

I pre-process all off-line video with Avisynth then, for H.264, I prefer 
MainConcept's reference encoder - although it doesn't allow the sort of 
tweaking I like to do when the source or the target profile video needs 
help. I will often use x264 if I need to really attend to the quality of 
the target output profile.

H.264's CABAC, Trellis and In-Loop filtering should be used if you need 
to keep bitrate low and want to retain higher quality at the expense of 
higher decode complexity.

If encoding time isn't a big issue, I recommend multiple passes if you 
need to keep the filesize/bitrate at a specific level. If you want a 
specific quality and size or bitrate isn't an issue; then go for 1-pass CBR.

One note about DV25--->MPEG-2--->H.264 encoding path is the color-space 
loss. As NTSC DV uses 4:1:1 color-space there's an issue with the 
conversion to the 4:2:0 color-space used in MPEG-2. There a differences 
you should look out for in most common DV decoders.

1.) The Canopus decoder performs the RGB conversion it seems to use a 
luminance range of 16-235 instead of 0-255 used normally for computers. 
It does not seem to apply any sharpening of the decoded image (which is 
good in my opinion). It supports YUV2 and RGB color space. The biggest 
problem is the known, but never fixed "chroma mapping" bug for both YUV2 
and RGB. The only way I know to fix it is with Avisynth's 
"Reinterpolate411" function here, because it reinterpolates the 
duplicated chroma lines (duplicated by the codec in YV12->YUY2)

2.) MainConcept's DV decoder sharpens the video a bit too much for my 
taste - this can increase noise in the subsequent transcoding. In both 
RGB and YUV2 modes it uses a luminance range of 0-255.

3.) Microsoft/Panasonic's DV decoder: Similar chroma bug as Canopus. But 
for those using Avisynth, use the "FixBrokenChromaUpsampling" function

4.) Cedocida (http://www.student.uni-kl.de/~dittrich/cedocida/) set to 
DV YV12 output for transcoding to H.264 (no color-space changes) 
produces the best output in my opinion.

Sorry for the scattered commentary - I don't know Frank's workflow so 
I'm just sharing some lessons learned the hard way.

Rich Griffin 
()
(216) 589-9626 ext. 158
OSTN


Frank Fulchiero wrote:
> Thanks everyone.
> I'm still learning about H.264 encoding at lower bit rates, and the QT 
> Broadcaster stream is coming from a DVD changer (DV>MPEG2>MPEG4), so 
> hopefully quality will improve. Some of the material was from VHS...
> Frank
>
> On [Aug 8], at 9:59 AM, Tim Chown wrote:
>
> > On Fri, Aug 08, 2008 at 09:52:23AM -0400, Frank Fulchiero wrote:
> > > Hi all,
> > >
> > > I made a few changes in our network parameters.
> > > Can someone outside the state of Connecticut please tell me if they
> > > see the Connecticut College program group in VLC, with three
> > > programs, and that one of the streams work (don't have to check all 3)?
> > > Still tweaking H.264 encoding, so one of the streams may crash on VLC
> > > on OSX.
> > >
> > > Sorry to trouble you.
> > > If you answer to the list also, then all I need to bother is one
> > > person....
> >
> > Hi,
> >
> > I see three entries in SAP for Connecticut College here in England.
> >
> > All three play.   Subjectively the quality is quite poor.    None of
> > the streams appear to crash OSX.
> >
> > --
> > Tim