Xvid play better with Divx decoder?

[miniyoyo]

I found that Xvid movie is playing better(color, lightness and responsiveness to player control) with the Divx decoder(mine is Divx Pro 5.2.1). It is strange that with the original Xvid decoder(1.1.0beta bundle with Autogk 2.1.8.1beta), the play back is not as good as Divx decoder. Anyone tried and can share your experience about that?

[JoeBG]

DivX Decoder has "build in" Postprocessing, that´s all.

[Teegedeck]

Yes, and if you use ffdshow to decode XviD you can modify the picture in any way you like; postprocess to get rid of blocking; increase contrasts, colours, brightness, sharpness; add film-grain to encodes with a too-artificial look... You see, it's nothing special to make an encode look 'better' through the decoder. XviD's own decoder also can do some of those things in case you didn't notice. But the decoder should also be capable to pass the picture through just 'as it is'. Actually, that is the original purpose of a decoder.

[miniyoyo]

So I mean "Divx decoder with postprocessing" playback Xvid movie smoother than "Xvid decoder without any postprocessing". Suppose Divx decoder with postprocessing cost more resources, right? This is the point. At least Xvid decoder should playback Xvid movie smoothly.

[celtic_druid]

On my system XviD's decoder uses maybe 5% more CPU than ffdshow. I don't have DivX's decoder installed, but somehow I seriously doubt that DivX's decoder with pp uses less CPU than XviD's.

[IgorC]

FFdshow has most optimal decoder for Xvid. Less CPU usage and very flexible
settings for postprocessing mode : luma, chroma, deringing, brighnest, sharpness etc.

[non@me]

If your movie is encoded using GMC divx decoder docodes for only 1 warppoint. But XviD has 3 warppoints.

[celtic_druid]

I seem to recall reading 3 wp GMC decoding support in a DivX changelog ages ago. Besides if you could only decode 1 out of 2/3 and it looked ok then why wouldn't they do that on hardware standalones?

[videomixer9]

If it comes to postprocessing capabilities I have to say that DivX still seems to be far superior over ffdshow or the xvid deblocker which comes with the decoder. DivX Decoder actually manages to postprocess the video without losing too much details. With ffdshow postprocessors I noticed that all of them are washing out the picture or even produce flickering parts in the picture with accurate deblocking. While the DivX postprocessor had some dotting effect the ppl complain about in earlier version that problem is fixed, the problem nowadays seems to be still in ffdshow or xvid when deblocking, it's just not a dot but a whole area being affected. ffdshow or xvid are imo plain smootheners and not real deblockers, their affect on details is just plain horrible, divx sometimes rather keeps a block or two to not lose the details, faces and small structure being all washed out is no postprocessing for me but ruining the video, DivX manages to actually reduce blocks while keeping the detail and colors, I wonder when someday some of those developers that work on those deblockers for mplayer etc. fix their lousy postprocessing, but as they either don't use it or think it's fine, which it isn't (well many ppl even think that this washed out shit is good postprocessing), they won't prolly change a thing. And this "DivX is commercial shit now" thinking doesn't change the fact that their postprocessor is superior.

[SeeMoreDigital]

XviD's revised DSdecoder (as provided with Celtic Druid's) new XviD Beta2 builds, I think is very good now....

http://img366.imageshack.us/img366/7955/xviddsdec8tk.png

And given that XviD's "XviD.ax" file only comes in at just 80KB. It compares remarkably well to DivX's enormous "DivX.ax" file!

But I guess I have to ask.... As MPEG-4 decoder filters go.... What other tweaks could be added to XviD's decoder filter?


Cheers

[temporance]

If it comes to postprocessing capabilities I have to say that DivX still seems to be far superior over...@videomixer9: thanks for the annecdote. I'm actually interested in finding out more about ASP postprocessing: how much difference it can make and which implementation gives best results (perceptual and metric). I'm trying to collect information here: http://forum.doom9.org/showthread.php?p=727749 so if you could add anything there, that would be great.