Help with Simpsons / Animation...

[teknomedic]

Rather than spill my guts again... here's a link to a thread I started with all needed info I would think.
http://forum.doom9.org/showthread.php?t=166875

Long story short is that even after IVTC, I'm still getting random interlaced type distortions in my videos. From my google searches I've learned that this generally affects Simpsons, Futurama, etc and has been discussed by various people in various threads for several years (early 2000's), yet I can't find a single guide or "good" information on how to fix it.

Thank you for any help.

[teknomedic]

I'm calling my problem fixed at this time... just follow along with the other thread to understand why and how.

Unless someone can offer more advice on doing this "correctly"???... in the end though... I have the results I was looking for with no noticable change in visual quality (and no interlacing issues) using the "motion compensation" option.

[jagabo]

I only looked closely at the first sample, the one from the title sequence. I saw two problems which make it impossible to perfectly IVTC.

At 7.0 to 7.7 seconds, while zooming into the letter P of "SIMPSONS" the telecine pattern of the outer title and the hole in the P don't match. Any field matcher (like TFM()) will be confused by this.

At about 14 seconds in to the shot the video was slowed down with field blending (after the video was telecined). This causes fields to have a double exposure look.

You could use a script like:

Mpeg2Source("Sample01.demuxed.d2v", CPU=6, Info=3) 
TFM(d2v="Sample01.demuxed.d2v") 
TDecimate() 
VInverse()

VInverse() gets rid of most of the residual combing from the TFM().TDecimate() sequence. Or you could try:

Mpeg2Source("Sample01.demuxed.d2v", CPU=6, Info=3) 
QTGMC(preset="fast")
SelectEven()
TDecimate()

[teknomedic]

wow... thank you looking deeper into the samples. This explains a lot of the "strangely telecined" remarks I was reading on google searches.

hmmm... I'm certainly willing to try the scripts, but I think I'll still have the same basic issue explained in my final post... I can't seem to get MediaCoder to use AviSynth script AND encode directly from the VOB (error 13).... and if I convert to another container while applying the AVS script first... I get strange results when I then try to encode to my final H264/MP4 setup.

Any thoughts on that?... or perhaps another program to try with VOB source and AVS script??

Thank you again.

[poisondeathray]

for avisynth, use dgindex to index the vob, then mpeg2source() as in jagabo's script

[teknomedic]

Ok.. so I get a D2V file after indexing... then what? I've tried different variations using the AVS and D2V file in both MediaCoder and VirtualDubMod... nothing but errors in both. VituralDub gives the error "there is no fuction named Mpeg2source" Meh.

[jagabo]

Put a copy of DgDecode.dll in AviSynth's plugins folder.

[teknomedic]

lol.. that got me to line 4... now "there's no function named Vinverse" ...assume I need another mystery.dll ??

[jagabo]

Yes, you need vinverse.dll in AviSynth's plugins folder. http://avisynth.org/mediawiki/External_filters

[teknomedic]

I appreciate the crash course... I appear to finally be directly encoding the VOB using AviSynth script via MediaCoder. So... guessing ATM that it was the missing DLL files I didn't know about.

Will take a small bit to see resulted encode... crossing fingers this will do it.

"IF" this works... how would I make a general script that would work on a list of say 20 VOBs?

Mpeg2Source(“$(SourceFile)”, CPU=6, Info=3)
TFM(d2v="Sample01.demuxed.d2v")
TDecimate()
VInverse()

I think I have line 1 correct... it's line 2 I'm not sure off... would this work??
TFM(d2v="$(SourceFile).d2v")

[teknomedic]

Well... lost audio and now instead of interlacing errors I have odd looking super dark but still kinda transparent spots that are now where the interlaced errors used to be. hmmm... I'll try the second script you offered.

[jagabo]

Mpeg2Source() only includes the video. You have to mux the audio separately. Normally you just take the AC3 file that DgIndex created and mux that into your final product.

[teknomedic]

Sorry about delay... fell alseep.

Well... that explains the audio then, lol.

So here's the same frame of the same scene...
Picture left is MediaCoder, IVTC Primitive w/ "fast" Motion compensation
Picture right is using either of the scripts provided.



I do get dark spots using the settings in MediaCoder too (just not as pronounced or frequent) similar to using the provided AviSynth scripts. When I view the VOB in the same frames I can see and understand why the spots appear (probably no way to get rid of them entirely) as it has to do with some odd interlacing transparency during what appears to be a blended frame?

I do have to give a nod to the AviSynth scripts producing a slightly cleaner image overall though when viewed in motion... but it's lost due to larger more pronouced dark spots.

The AviSynth script also seems to produced a slightly out of ratio horizontal stretch too.

hmmm... still thinking the pre-built settings in MediaCoder are winning the trials. Thoughts?

[teknomedic]

One more example with VOB center...

----Left (top) *first field*
----Left (bottom) *second field* -- (MediaCoder, IVTC, Motion Compensation)

----Center VOB -- (ripped from DVD, no changes or filters, showing both first and second fields interlaced)

----Right -- (AviSynth scripts with "blended?" first and second fields??)




Again, I do get some darkened areas when I use the MediaCoder settings too... just not as bad.

[jagabo]

dark spots

Was that frame in the samples you uploaded? I did see some problems where interlaced YV12 was mishandled by the previous encoding leading to blended chroma channels. It think that's the problem with the dark spots. You might try using SelectOdd() instead of SelectEven() in the second script. But it may just move the problem to different frames.

When TFM() can't find complimentary fields it uses a blend deinterlace on the remain comb artifacts. If it doesn't see the comb artifacts (which sometimes happens with very small ones) VInverse() will blend them.

You can add the argument "pp=0" to disable TFM's post processor. That way it won't attempt to remove the remaining comb artifacts. Also remove VInverse(). That will let you see combing from orphaned fields.

The AviSynth script also seems to produced a slightly out of ratio horizontal stretch too.

AviSynth doesn't pass the aspect ratio information to the encoder. You must specify the PAR or DAR manually in the encoder.

[teknomedic]

I appreciate the help and I've learned a lot.

Those frames weren't in the samples... this was from encoding the full episode and this scene was one of the more pronounced "dark spot" areas.

Didn't know about the aspect ratio of AviSynth PAR/DAR settings... good to know.


Your explaination explains about as much as I understand at this point plus some about the dark spots... but it makes sense. After reading your description it also explains why I still see those spots in lesser amounts using the pre-built MediaCoder settings too.

With some more tweaking I think I'll get it.


Now then... onto cleaning up a PAL dvd rip of Family Guy, haha.

[jagabo]

If you upload a sample of the source that shows that problem I'll take a look at it.

[teknomedic]

Here ya be...
http://www.mediafire.com/?o9cbw77h2a3yfdj

[jagabo]

The problem is in your source video. As I noted before, interlaced YV12 has interlaced chroma channels as well as an interlaced luma channel. Some program treated (or the video was improperly filtered) and the chroma channels were treated as if they were progressive. This has caused the colors from the two fields to blend together. I think it's just an accident that some processing sequences manage to eliminate more than others.

[poisondeathray]

I think it's just an accident that some processing sequences manage to eliminate more than others.

I agree ...


@teknomedic

The problem fields are grouped too close together in that sequence with mr. burns that manual decimation with tdecimate override file or yatta won't work either. You would have to manually replace with duplicates (e.g. freezeframe) or interpolate frames

[teknomedic]

I don't disagree with your findings.

However, I don't plan to spend any more time trying to fix a few frames. I really do appreciate the help... thank you!!... also... I mean no disrespect, but I basically went in one big circle yesterday trying to fix this.


Using MediaCoder with IVTC set to "Primitive" and Motion Compensation set to "fast" has resulted in getting me 98%-99% of what I need... and honestly, it only took checking a box and adjusting a pull down menu vs writting scripts and installing more software to learn.


On the plus side I learned a lot more about AviSynth and MediaCoder (as well as a few other programs) not to mention what I learned about Telecine and Chroma etc... so it's not a total loss.