Codec review and analysis - any ideas?

[javipas]

Hi all,

I'm working on an article that pretends to review and analyse the
most famous and best A/V codecs out there. I wonder if you have
any suggestions that may help me complete the task. The issues
I'd like you include your ideas are:

Video codecs/containers: AVI/ASF, WMV, MPEG-1,2,4, RealVideo,
QuickTime, Matroska, VFW, H.264, DivX 3.x/4.x/5.x, XviD,
3ivx, XVD, MP4, OGM, XCD/VCD/SVCD/cVCD

Audio codecs/containers: MP3, Ogg Vorbis, WMA, RealAudio,
AAC, Musepack, WAV, MPC, Liquid Audio, MPEGPlus, VQF
Monkey’s Audio, AC-3.

I did some investigation on this subject a few months ago, but only
Related to audio codecs. I ripped a song from an AudioCD and
compressed it in various formats. Then I compared the original
waveform to the new ones using CoolEdit’s oscilloscope…

I understand this is a good way to test the codecs, but I’d like
to know if there is some “scientific” way to test the video codecs.
Is there any computer software that can be used to do this? The
Only way I see to get the differences is to analyse one still image
Of the video… but how to analyse the final waveform of the video
Streams….

I see there has been a good video codec comparison on Doom9 on
http://www.doom9.org/index.html?/codecs-203-1.htm. It has good
points, but I’d like to get more detail to recommend the codecs as long
as they can be used to:

- Compress DVD-Video
- Video-Conference
- Email (broadband)
- Email (56 kbps)
- Streaming servers (high bandwith for broadcasts)
- Idem, with limited bandwith

I guess this is pretty tough, but I’d like to complete a good review,
And I know for sure there is people over here that knows a lot far
More about A/V codecs that I currently do!! :-D

Waitinf for comments,

Greetings

Javi

[bond]

Originally posted by javipas
I did some investigation on this subject a few months ago, but only
Related to audio codecs. I ripped a song from an AudioCD and
compressed it in various formats. Then I compared the original
waveform to the new ones using CoolEdit’s oscilloscope…so your only comparing the codecs graphically?
have a look at the hydrogenaudio forum, there are lots of discussions on audio codec comparisons (but beaware you would get bashed hard there if you state graphically comparing audio codecs is a good thing ;) )

actually most think that double blind group listening tests are the way to go

I understand this is a good way to test the codecs, but I’d like to know if there is some “scientific” way to test the video codecs.
Is there any computer software that can be used to do this? The
Only way I see to get the differences is to analyse one still image
Of the video… but how to analyse the final waveform of the video
Streams….i dont think that there is THE scientifc way to compare video codecs

sure there are some quality measurement indeces available, which are often used to compare codecs
but still the truth (meaning if you look at the video yourself) often prooves them wrong

maybe the most scientific way to compare video codecs is how mpeg does it, i am sure if you search for it you will find some documents on how they do it

[javipas]

In fact, I've been reading some post on the hydrogen forums
and I can see that graphs aren't the best way to do it. I
will use WinABX to make blindtests and see what happens. I'm
surfing as a pro right now, there are tons of info, so any
good link about the issues I commented would be useful. I have
yet like 50 good appearing links to read.

Thanx for the quick comments!

Javi

[Selur]

you might also want to check out:
http://www.everwicked.com/vqstudio/
and:
http://www.everwicked.com/content/QualityMetrics/

Cu Selur

[javipas]

Both seem helpful, thanx selur.

Keep on investigating... tons of reviews about the
subject, it seems there's no end on this!

[On2Tech]

Originally posted by javipas
Hi all,
Video codecs/containers: AVI/ASF, WMV, MPEG-1,2,4, RealVideo,
QuickTime, Matroska, VFW, H.264, DivX 3.x/4.x/5.x, XviD,
3ivx, XVD, MP4, OGM, XCD/VCD/SVCD/cVCD


Javi

I'd like to suggest that you include VP62 in your comparison. I think your results will be more complete and also that VP62 should perform quite well against the other competitors you listed.

Sagittaire's Comparison (http://forum.doom9.org/showthread.php?s=&threadid=64789&highlight=sagittaire+vp6)

P.S. I'm not exactly an impartial observer....

[Selur]

yup, the on2tech codec: VP62 should be part of it, depending on when you plan to do the codec ogg theora might also be a nice addition.

Cu Selur

[SeeMoreDigital]

I don't know about some of you guys but I've found that certain codecs work much better than others when encoding at larger (ie: 720x576) frame sizes.

I guess many people on the forum use Gordian Knot to generate their Mpeg4 (DivX, XviD) encodes, after having performed a compressibility test to determine the ideal image pixel frame size that's required.

However, I've found that codecs such as RealMedia, WMV9 and VP6 don't seem to be so affected when encoding at bitrates below 1000kbps and still manage to look very good when using larger image pixel frame sizes - ie: more pixels.

All of which I think should be considered. As not everybody watches their encodes on PC monitors. Some of us watch them on large screens!

Cheers

[javipas]

My idea is to include as much codecs as possible, though as
many of you would know, taking all of them would be as
useless as painful, so the idea is take a subset of them and
see how do they perform. On2Tech, the tests seem pretty detailed,
so i will take good note of them.

As for the large resolution, I was a year ago in a Microsoft
Mobile Developpers Conference and they exposed in one of the
conferences the specs for low and high transfer rates (telephones,
HDTV), and WMV seemed (of course) to perform great on both
scenarios. I haven't tested that myself, but if you have experiences
with this scenario as well... (anyone encodes 1080p?)

Bye!

[SeeMoreDigital]

Originally posted by javipas
... I haven't tested that myself, but if you have experiences
with this scenario as well... (anyone encodes 1080p?) I've done some tests from 1080i 16:9 HDTV sources. But I ended up, de-interlacing them before converting them to an 'anamorphic' frame size of 576wide x 720high.

Despite the fact they were not encoded using 1280wide pixels they came out very well. Much better than expected!

Cheers