DivX 5.03

RadicalEd0 · Post by **RadicalEd0** » Sat Jan 25, 2003 2:06 pm

http://www.divx.com/divx/index.php

Doom9 wrote:It supports interlaced video, has a new post-processing algorithm which should improve perceived video quality during playback, has a video buffer verifier model in both one and two pass which can be used restrict the maximum peak bitrate and ensures that the encoded stream never violates the buffer of an MPEG-4 compliant decoder. Furthermore there's Nth-Pass encoding which means you can now perform more than 2 passes, the motion estimation algorithm has been optimized for higher resolution, the MP4creator has been removed for compliance issues but should be back in a later release, along with the intelligent IVTC functionality and some issues with DivX3.11 compatibility have been fixed

Ashton · Post by **Ashton** » Sat Jan 25, 2003 4:47 pm

So what's your call, Ed? Yay or nay?

Brolli411 · Post by **Brolli411** » Sat Jan 25, 2003 5:16 pm

Quality's great, just on all DivX encoded videos, in the bottom right hand corner, there's a DivX logo.

RadicalEd0 · Post by **RadicalEd0** » Sat Jan 25, 2003 6:09 pm

pfft the decoder places a DviX Video logo every ones in a while over the video. Filter Properties/Quality Settings/Disable Logo might possibly solve your problem :\

Well I'm testing it now, I'm not really interested in the quality of the codec itself, it was the thing about the decoder having higher quality postprocessing that I figured would be a good thing in the amv community, since most people use the divx decoder but alot dont use divx itself.

I did 3 passes so far and judging from avisynth's subtract function there is a huge difference between the second and third pass.
I'm not overly impressed with the quality so far but I have to do an xvid encode of the same thing and see how it compares yet

RadicalEd0 · Post by **RadicalEd0** » Sat Jan 25, 2003 6:13 pm

once in a while* >_< damn homonyms or whatever

Knowname · Post by **Knowname** » Sat Jan 25, 2003 7:01 pm

RadicalEd0 wrote:once in a while* >_< damn homonyms or whatever

dam now I'm gonna have problem with those... Never had problem with their there and they'res till somebody brought it up now this... dam you Ed0~!

RadicalEd0 · Post by **RadicalEd0** » Sun Jan 26, 2003 2:33 am

ok, tests are finished.

codecs tested:
DivX 5.03: 2 and 6 passes - 13,658 kb and 13,666 kb respectivley
XviD (latest Koepi cvs snapshot): 2 passes - 13,956 kb
3viX D4: only has one pass - 14,332 kb
DivX 3.11 SBC: 3 passes - 14,358 kb
Windows Media Video 9: 2 passes - 14,339 kb
RealVideo 9: 2 passes - 14,630 kb

clip used: frame 1307 - 3870 of Mission Improbable. 24fps, 512x384, 1024 kbps vbr used in all cases
both virtualdub and codecs with respective decoder filters set to max post processing for analysis

DivX 5.03, both 2 and 6 passes, actually surprised me. There was significantly less blocking and ringing than XviD in Vdub, instead areas which would have blocked took on an almost postprocessing esque muddiness. However it did look better than XviD's blocks. With full post processing however XviD caught up and was on par with the 6 pass DivX encode.

I was a little dissapointed in XviD in that as I mentioned there was some pretty noticable blocking and noise artifacts. Post processing cleared them up, but, I'm not in any condition to go around using 100% post processing on my 1ghz machine for normal viewing.

3vix started out good but got progressivley worse. Eventually it turned into an all out block fest that was uglier than the XviD by a good amount. Postprocessing did a good job of getting rid of the blocks, but also made the image kind of screwy, as in, not much better than blocky artifacts.

DivX 3.11, not as much noticable blocking as XviD or 3viX here, however it made up for that in a lot of mosquito noise. Postprocessing did an OK job of blurring this out but still wasnt great overall.

WM9, for all the claims made by M$ about this I'm not really impressed. Since WM is directly postprocessed, the only comparison was with the other codec's postprocessing, in which case it was sharper but also noticably blockier than XviD postprocessed. It holds a candle to Real9 in that it is a bit sharper but again it loses out in that it still has significant blocking.

RV9, I know I'm a RM whore, but honestly and objectively Real did do the best job here. Much sharper than XviD postprocessed, but with no noticable blocking or ringing. Slightly sharper than DivX 6pass with post, yet still with less leftover artifacts. Real stuff always looks sort of like its been run through 2dcleaner, as opposed to just being blurred like most postprocessing does, which is in part why it works so well with anime.

Overall I'd have to rank it like this:
1 - RV9
2 - DivX 5.03
3 - XviD
4 - WMV9
5 - DivX 3.11
6 - 3viX

not surpisingly, 3viX is last, and RV9 is first. The defunct DivX 3.11 is second to last. The upsets are in between. There was no fine line between divx 5.03, XviD, and wm9, it came down to one shot where a plane of color was more noisy in WM9. XviD came in after DivX due mainly to the excess blocking out of post. ffdshow's post turned out much worse than DivX 5's, with it xvid was under wm9. However with the divx 5 decoder 5.03 and xvid were very very similar.

So basically RV9 still rocks, XviD is still l33test, DivX 5.03 is a bit better quality than XviD, 3vix and WMV still suck and divx 3.11 is still old

Well I hope this info is useful somehow to someone because it took me like a fucking hour and a half to write it all :/

RadicalEd0 · Post by **RadicalEd0** » Sun Jan 26, 2003 2:47 am

I forgot to mention no qpel or bframes or anything were used unless wmv and rm automatically use those which would suck but :\

AbsoluteDestiny · Post by **AbsoluteDestiny** » Sun Jan 26, 2003 3:44 am

Did you notice any actual difference between the nth pass divx and the 2nd pass? I'm guessing nth pass encoding is a bit overkill - why can't they just improve the curve calculations before the 2nd pass?

The results seem about right, but the thing I guess I'd have more objections to is the RM test. You say that RM looks like it's used 2d cleaner - well, it has basically. As you know, RM has noise reduction filters built into the codec whether you want them or not. Personally I'd rather choose my noise reduction.

Also you can't beat xvid for compatibility when encoded as ISO compliant

trythil · Post by **trythil** » Sun Jan 26, 2003 5:11 am

Could you provide frame grabs from each of those test encodings?