|
|
| i4004 |
| Posted: Mar 21 2004, 09:05 PM |
 |
|
Advanced Member
Group: Members
Posts: 2432
Member No.: 4935
Joined: 24-June 03
|
"hand" sequence rides again... 
we know who wins on lower bitrates,but what's with higher bitrates...
http://kotisivu.mtv3.fi/ii00i/5Mbit/5mbit.html
hmm..i should give the encoding times too:
Nandub-4'09"
Tmpgenc-7'05"(normal motion search precision)
CCE-3'43"
MEncoder-3'40"
QuEnc-4'09"
CCE is somewhat faster in CBR mode(as opposed to VBR mode) and MEncoder is a bit faster in VBR mode,but MEnoder in CBR mode is still faster than CCE in CBR mode...all times above are for CBR modes..
MEncoder was set to use same quantization matrix as CCE and Tmpgenc use....
here are the versions;
CCE SP[SSG] 2.66.01.04
TMPGEnc Plus 2.5[2.59.47.155,Core 1.92.142]
MEncoder dev-CVS-040229-14:06-3.2
QuEnc 0.45
here's the mencoder batch used;
| CODE | @echo off
set outputdir=D:\Video\Finished
set fakeavi=fake.avi
set mpg=video.mpg
set intra="8,16,19,22,26,27,29,34,16,16,22,24,27,29,
34,37,19,22,26,27,29,34,34,38,22,22,26,27,29,34,37,
40,22,26,27,29,32,35,40,48,26,27,29,32,35,40,48,58,
26,27,29,34,38,46,56,69,27,29,35,38,46,56,69,83"
set inter="16,17,18,19,20,21,22,23,17,18,19,20,21,22,23,
24,18,19,20,21,22,23,24,25,19,20,21,22,23,24,26,27,20,
21,22,23,25,26,27,28,21,22,23,24,26,27,28,30,22,23,24,
26,27,28,30,31,23,24,25,27,28,30,31,33"
mencoder -of mpeg -ovc lavc -lavcopts vcodec=mpeg2video:
intra_matrix=%intra%:inter_matrix=%inter%:mbd=0:vbitrate=5000:
vrc_buf_size=1835:keyint=15:vmax_b_frames=2:vqblur=0:aspect=4/3
%mencpath%\%fakeavi% -o %outputdir%\%mpg%
pause |
settings used for CCE and Tmpgenc are same as used in this
http://free-zd.hinet.hr/ikostic3/laced.htm
test,but interlaced encoding was OFF as this was progressive sequence...(kerneldeinterlace used),and bitrate was adjusted...
QuEnc used hi-quality mode....
filesizes for this and interlaced tests;
http://i4004i4004.bizhosting.com/filesizes/
--------------------
my signature:
 |
|
| kurtnoise |
| Posted: Mar 23 2004, 09:12 AM |
|
|
Moderator of the Vdub Support Forum
Group: Moderators
Posts: 265
Member No.: 1206
Joined: 6-December 02
|
Hi i4004 !! 
Could you add the QuEnc from Nic in your tests ??
10x.
--------------------
Visit the French United Video Forum : http://www.unite-video.com
|
|
| i4004 |
| Posted: Mar 23 2004, 04:02 PM |
|
|
Advanced Member
Group: Members
Posts: 2432
Member No.: 4935
Joined: 24-June 03
|
sure.....hand sequence won't be erased soon,so anything goes....
my god,this is awesome test sequence;even at 5Mbit/s many systems have problems reaching what mpeg calls "transparency" on this sequence.....
tmpgenc will transfer some frame portions better than the rest,while it'll make other things worse...etc.
anyway,quenc coming..i expect it'll be nearest to the mencoder.....
--------------------
my signature:
|
|
| i4004 |
| Posted: Mar 24 2004, 03:19 AM |
|
|
Advanced Member
Group: Members
Posts: 2432
Member No.: 4935
Joined: 24-June 03
|
added QuEnc ,mencoder batch,encoder versions....
filesizes for this and interlaced test....
--------------------
my signature:
|
|
| kurtnoise |
| Posted: Mar 24 2004, 09:43 AM |
|
|
Moderator of the Vdub Support Forum
Group: Moderators
Posts: 265
Member No.: 1206
Joined: 6-December 02
|
Great....so, just one more question : why did you use specifically the SBC for the MPEG4 ?? It's your favorite ??
10x
--------------------
Visit the French United Video Forum : http://www.unite-video.com
|
|
| i4004 |
| Posted: Mar 24 2004, 04:47 PM |
|
|
Advanced Member
Group: Members
Posts: 2432
Member No.: 4935
Joined: 24-June 03
|
yes...my tests revealed some nasty problems when divx4/5 and xvid do a noisy stuff so i don't use those...
you can read more here
http://forum.doom9.org/showthread.php?threadid=71485
actually,when fcchandler says that "mpeg2 looks sharper than mpeg4" i think he's saying "mpeg2 looks sharper than divx5 and xvid".....
--------------------
my signature:
|
|
| fccHandler |
| Posted: Mar 25 2004, 04:49 AM |
|
|
Administrator n00b
Group: Moderators
Posts: 3961
Member No.: 280
Joined: 13-September 02
|
Well, my opinion hasn't changed. MPEG-1 and -2 are older and simpler schemes. MPEG-4 uses some advanced ideas to achieve its higher compression ratio, but at the same time it hopes you won't notice the subtle ways it degrades the video. I've recently come across a few sources that look ugly in MPEG-4 no matter how high a bitrate I choose.
I really believe that "transparency" is impossible with MPEG-4 at 5mbit/s or even at 50mbit/s. I think MPEG-2 is more faithful in its reproduction of the video (and yes, that includes sharpness), but as a general rule it does require more bits, and more time to compress...
--------------------
May the FOURCC be with you... |
|
| i4004 |
| Posted: Mar 25 2004, 11:03 PM |
|
|
Advanced Member
Group: Members
Posts: 2432
Member No.: 4935
Joined: 24-June 03
|
| QUOTE | | I've recently come across a few sources that look ugly in MPEG-4 no matter how high a bitrate I choose. |
i generally use such sequences for testing....(ie. hard to encode)
but.....i had a breakthru last night.......(it'll sound weird,i know)
i was able to make a 700kbit/s mpeg1 encoding that looked better than mpeg4 on slightly higher bitrate!!!
(hand sequence scaled to 384x288@25fps..still a bitch to do properly...divx3 flopped like i never saw it before!...770kbit 384x288 mpeg4 is the stuff i normally do for a particular type of content...)
i couldn't believe my own eyes!
but..rather than being overwhelmed with two tests i'll do some more (on different resolutions) and we'll see what happens...if i get consistant performance from mencoder,it seems i may as well convert to mpeg....
(i can only guess what has happened;b-frames and excellent motion estimation have shown how close mpeg4 really is to mpeg1/2.......tmpgenc has "high motion search precision"..mencoder simply has fast and excellent motion estimation.....mencoder's 720x576 mpeg2 interlaced encoding is mind blowing!)
you will remember that in my previous test mpeg1 looked sharper than mpeg4 on 448x336@770kbit/s,but there were some "chroma smearing/blocking " issues on mpeg1 clip (although my "hand" test clip is semi-synthetic too,as large portions of the image are computer graphics..a thing for which mpeg1/2 were never designed....)
look at me,i did a full circle....mpeg1/2,mpeg4,back to mpeg1/2...hehe
anyway,i'll keep you informed;i don't care which system is used or how i call it;i only need sharpest image possible!(ok,it needs to be fast on encoding too...hehe)
i will dump mpeg4 in a blink of an eye if mpeg1 can beat it!worry not!
did avery said that he uses mpeg1 too? ( )
enough if silly chit-chat...gotta do more mencoder stuff....must see if this was for real or just a dream i had yesterday...
--------------------
my signature:
|
|
| fccHandler |
| Posted: Mar 26 2004, 06:44 AM |
|
|
Administrator n00b
Group: Moderators
Posts: 3961
Member No.: 280
Joined: 13-September 02
|
Just as an example, one of my favorite codec-killing scenes is the first appearance of Darth Vader in Star Wars (that is, episode 4). He steps into a corridor which has just been cleared by stormtroopers, but the lingering haze of smoke in the scene ruins every MPEG variant I've tried; even MPEG-2 has trouble with this scene.
My source is a THX-certified VHS tape of the "original version" of Star Wars, released in the mid-90's IIRC. Needless to say, I'll be curious to examine this scene when the Star Wars DVDs are released in September (keeping my fingers crossed for that).
--------------------
May the FOURCC be with you... |
|
| i4004 |
| Posted: Mar 26 2004, 07:18 AM |
|
|
Advanced Member
Group: Members
Posts: 2432
Member No.: 4935
Joined: 24-June 03
|
i uploaded this example (the one i mentioned above) as i know you'll think i'm crazy untill you see it for yourself...
http://free-st.htnet.hr/hsari11/clips/divx...3-300frames.avi
http://free-st.htnet.hr/hsari11/clips/mpeg1-12sec.mpg
(right click,save target as....)
further tests didn't went so well for mpeg1;
512x384@1,2Mbit divx3 won(noisy broadcast source)
384x288@750kbit divx3 won again(marx brothers 30's movie "monkey business",clean broadcast,but the movie itself is old....)
i also compared mencoder's mpeg1 to mencoder's mpeg4 (on 512x384) and i didn't saw difference....they both look slightly less sharp than divx3....(ie.same stuff as i experienced with ffvfw mpeg4 versus divx3 on lower bitrates...)
so it seems as if divx3 hated this particular sequence(more than mencoder)...
yeap.....it was only a dream....
(btw. i succesfully solved the gradient blocking(wall in the background) with mencoder's scplx_mask=0.5...but....that scene was ok with tmpgenc too,and your star wars scene probably wasn't..heh...)
--------------------
my signature:
|
|
|
