Nasty COMP effects in XviD --- Please HELP!

[kitanai]

I am a newbie to XviD and came across an nasty problem when checking the result of my encoding. :confused: Here is what happend, I hope someone will find my error:

I tried to encode a DVD rip, which is originally in NTSC (DVD2AVI says INTERLACED, so it did not check FORCED FILM) and in 4:3. I think it was originally recorded on camera or tape and then converted to DVD (just to give you another hint).

For encoding, I used XviD 1.0 from Doom9 with the following settings:
- 2-pass
- Profile AS@L5
- Adaptive Quantization OFF
- QuarterPixel OFF
- B-VOBs -> Quantizer Offset 0.50
- PacketBitstream ON
- Motion Search 6
- VHQ 4

and in second pass Jawor-1CD-Matrix for Quantization Type.

In GordianKnot
- Resize Filter: Neutral Bicubic
- Field Operation was left NONE
- therefore, I left IVTC in avs -> correct Frame Count OFF

The Video looks great compared to my previous tries :D and can surely compare to the stuff an the web (many thanks to the developers of XviD!!!) However, I see these ugly COMP effects, especially in fast motion scene.

Ok, I read A LOT about IVTC and deinterlacing and the problems that DVD2AVI sometimes does not show what the source really contains --- I have to admit: I am still confused about it...

I am sure, someone more experienced can tell me which box to check to get the perfect result :p Many thanks in advance!

[manono]

Hello and welcome to the forum-

If you've done all your reading then you should be able to answer these. Is the source really interlaced, or only encoded as interlaced? Can it be IVTC'd or not? You said that you didn't use Force Film. You seem to say that you didn't perform IVTC. So did you deinterlace it? Ordinarily, if you don't Force Film, you do one or the other. Is it a movie? Ordinarily, if you don't Force Film, you IVTC movies. But that may depend on the master used for the DVD.

COMP effects=compression artifacts? What kind? Blockiness? Or are you seeing the thin parallel, horizontal black interlacing lines?

Maybe you can link to the DVD on Amazon.com or someplace. Maybe someone that has ripped it can help out if we know what the movie is. Maybe you can post the .avs you used, and the GKnot log.

And I think this has more to do with GKnot than with XviD, so maybe an XviD Mod can move it.

[kitanai]

Thanks for your quick response!

I will try to give you some more details on my issue...

- The source is on DVD, but originally was an NTSC tape (it's not a movie but was recorded on a digital camera, so it's not available at Amazon ;) )
- DVD2AVI shows : NTSC, 4:3, INTERLACED
- DVD2AVI also shows thin, parallel, horizontal lines on the egdes of moving objects (not necessarily during preview playback but when using the slider)
- these lines reappear in the encoded avi

I re-read the "IVTC & Deinterlacing tutorial" at doom9 and I think I have a pure interlaced source according to "Truly interlaced clips are stuff you capture with your camera, and TV shows originally shot for TV" and the fact, that VirtualDub show these lines in every single frame. :cool:

As far as I can conclude now, I am supposed to use "FieldDeinterlace" to get rid of the lines?!
What method can you recommend: FieldDeinterlace, FieldDeinterlace (no blend), Smart Bob...?

However, one statement still puzzles me: "when GK shows Inverse Telecine in Field Operation you have to check it". And GK does.
:confused:

I am sorry for asking these basic questions, but sometimes it is hard to find your way through all relevant stuff...

[Salseros]

The artifacts clearly show that you've forgotten to de-interlace, so at field operations choose de-interlace (the normal one). :)

I've encoded quite a few movies and have to say that I never had to use IVTC.

[manono]

Hi-

However, one statement still puzzles me: "when GK shows Inverse Telecine in Field Operation you have to check it".

That one's easy. If, when setting up your .avs in the Save And Encode box prior to encoding, you've checked "Inverse Telecine" under Field Operations, then in the Encoding Control Panel, when getting ready to encode, you also have to check the "IVTC in avs->correct Frame Count" box. But that evidently doesn't apply here, as you have interlaced content, and won't be performing IVTC.

What method can you recommend: FieldDeinterlace, FieldDeinterlace (no blend), Smart Bob...?

Try them out and see which you prefer. Open the .avs and where it says:

# DEINTERLACING (1)
#FieldDeinterlace()
#FieldDeinterlace(blend=false)
#TomsMoComp(1,5,1)
#
# DEINTERLACING (2)
#KernelDeInt(order=1,sharp=true)

Uncomment them (remove the "#") one at a time, save it, open it in VDubMod and see how the different ones look. KernelDeint is the SmartBob that GKnot uses, and the one I usually prefer.

Oh yes, I almost forgot. You have to also load the appropriate plugin. Up at the top, where it says:

#LoadPlugin("F:\DIVXST~1\GKnot\decomb.dll")
#LoadPlugin("F:\DIVXST~1\GKnot\KernelDeInt.dll")
.
.
#LoadPlugin("F:\DIVXST~1\GKnot\TomsMoComp.dll")

when using one of the FieldDeinterlaces, you uncomment Decomb.dll. When testing KernelDeint, you uncomment the KernelDeint.dll. And if using TomsMoComp, uncomment the TomsMoComp.dll.

Or you can just set it up from within GKnot one deinterlacer at a time, and then check it out in VDubMod, or run the Preview. But doing it in VDubMod allows you to advance one frame at a time, and study the results more easily.

[kitanai]

Many thanks for your help!!!

It was the FieldOperations-Option. Now the AVI looks great.

[kitanai]

Just to make sure, please let me post one more question...

What quality can I expect from my source?
That is, I resized it to 480x352 which sums up to 168960 pixles/frame. Audio was done in 96kbit/32kHz/VBR which is still ok.
All to fit a 700MB-CD-R (1h 47m).
GK calculates an avg. bitrate of about 800kBit/s (which ends up at 631305 kbytes in the XviD setting).
Of course, when in full screen playback it shows some quite visible artifacts...

On the other hand, I compared it to a movie from the net (576x240=138240pixels, 1h 46m), also a one-CD-XviD rip, which shows far less artifacts then my work.

Furthermore, it appears to me, that I do not seem to get better results even when switching to a 2-CD encoding; just more size, same picture quality.

Does this make sense... :confused:

[manono]

Hi-

You're comparing apples to oranges when you compare your source with a different one. Different videos compress differently. In addition, yours consists of more than 20% more pixels, and 25% more frames (if yours is at 29.97fps and the other one is at 23.976fps). So even if they did compress exactly alike, you'd still need more than 50% more file space to produce a result as good as the one you saw.

But that's what a compress test is for. Run one to see if your chosen resolution and file size is enough.

Deinterlacing is always an inferior solution compared to being able to Force Film or IVTC. That's why you have to be real careful to make sure that your source is interlaced to begin with.