chroma ghosting errors

[Katie Boundary]

https://scontent-lax3-1.xx.fbcdn.net/v/t1.0-9/13900362_1718006328463822_100809593258046277_n.jpg?oh=104f016156e24a883eafa3fb438ca78e&oe=582028E0

https://scontent-lax3-1.xx.fbcdn.net/v/t1.0-9/13700013_1711705509093904_8124036554805709401_n.jpg?oh=0b734d3ced517e0e4218927a71dd54e1&oe=581F2F9D

https://scontent-lax3-1.xx.fbcdn.net/v/t1.0-9/13718685_1711728372424951_7568096017440224120_n.jpg?oh=0b44dd0599626f498c8fe97b51849562&oe=581F66B1

Software used: DGIndex, AVIsynth, loadcplugin, ibob, Virtualdub.

The really weird thing is, I can't always replicate these errors, which means I can't pinpoint exactly where stuff starts going wrong. Has anyone else encountered errors like these? If so, do you know what's causing the problem?

[filler56789]

.........
Software used: DGIndex, AVIsynth, loadcplugin, ibob, Virtualdub.

and last but not least, the infamous JPEG compressor used by Facebook.

For the umpteenth time:

1) save your screen shots as PNG files

&&

2) use a decent image hosting service.

[Katie Boundary]

and last but not least, the infamous JPEG compressor used by Facebook.

For the umpteenth time:

1) save your screen shots as PNG files

&&

2) use a decent image hosting service.

The color distortions exist prior to exporting as still images, so it has nothing to do with png or Facebook.

[ChiDragon]

This is usually caused by improperly handling interlaced 4:2:0. For example, using ConvertToYUY2() instead of ConvertToYUY2(interlaced=true).

[Stereodude]

This is usually caused by improperly handling interlaced 4:2:0. For example, using ConvertToYUY2() instead of ConvertToYUY2(interlaced=true).
C'mon now... Don't encourage it.

[Katie Boundary]

This is usually caused by improperly handling interlaced 4:2:0. For example, using ConvertToYUY2() instead of ConvertToYUY2(interlaced=true).

So should I convert to RGB before doing anything else, just in case?

EDIT: Well, I tried that, and it completely screwed up the field order. I tried separatefields, then converttorgb, then weave, but that didn't fix anything. assume tff, assumebff, and complementparity all failed. So I stripped everything out and went back to having nothing but loadcplugin and ibob, and now I'm getting an all new error: interlaced chroma ghosting!

https://scontent-lax3-1.xx.fbcdn.net/v/t1.0-9/13902547_1718314435099678_1300812693360210537_n.jpg?oh=7ea464d26786d6b090b23ca8317191f3&oe=581E0D24

The gods have forsaken me :(

[manono]

So should I convert to RGB before doing anything else, just in case?
No, as the damage was (probably) already in your source. This isn't all that uncommon and TFM has a couple of settings that (sometimes) take care of it, chroma and mchroma. Or you can try bobbing followed by SRestore to the correct framerate to see if it takes care of it.

Perhaps if you posted a sample someone could have a look.

[Katie Boundary]

No, as the damage was (probably) already in your source.

I shall direct your attention to the following line in the OP:

I can't always replicate these errors

So, obviously not the source.

EDIT: it occurs less often when using Bob or Separatefields instead of ibob, and while I THINK I remember this error showing up with at least one of those two (South Park + separatefields comes to mind), I also have a tendency to go two weeks at a time without sleep, which can severely affect memory... so ibob is the leading suspect in this case.

[manono]

So, obviously not the source.

And we're just supposed to take your word for that? Where's your script and, more importantly, where's the requested sample?

[Katie Boundary]

And we're just supposed to take your word for that?

yes.

Where's your script

mpeg2source("101.d2v")

loadcplugin("c:\ibob.dll")

ibob()


and, more importantly, where's the requested sample?

The three that I gave in the original post weren't enough?

[filler56789]

...
The three that I gave in the original post weren't enough?

Nope, because manono requested video samples,
NOT horribly-artifacted JPG images.

P.S.: but I am very sure you already knew that.

[manono]

The three that I gave in the original post weren't enough?
Stop playing cute. Video sample. Untouched. Direct from the source. You can use DGIndex and the [ and ] buttons to isolate a small section that shows the problem after being filtered. Then File->Save Project and Demux Video. Make the resulting M2V available please.

And no one uses iBob. Except Katie Boundary, I suppose. Its only possible use is to get back the 'original' field after being filtered. And if you're trying to IVTC or otherwise make your source progressive, you plain just wouldn't use it.

[Katie Boundary]

Make the resulting M2V available please.

That's the part where we got stuck last time.

The key point here is that these errors cannot be reliably replicated, and that they only occur during deinterlacing. I was hoping that a problem with such unique characteristics would be well-known already, and that the cause would have already been identified. I doubt that you'd be more helpful if you had access to some untouched DVD footage just to confirm that this isn't a problem with the DVDs.

And no one uses iBob.

Well, if ibob is the source of the problem, then that might be why I'm the only one who's experiencing it.

[manono]

Then if you refuse to do as requested, there's nothing more to be done and no help to be provided. If you're asking for help, you post the script and a sample. Basic procedure.

If you're not asking for help then why does this thread even exist?

I'll be happy to delete this thread if you like.

[wonkey_monkey]

The key point here is that these errors cannot be reliably replicated

No, the key point is that we can't replicate the problem at all if you refuse a perfectly reasonable request for footage.

If your car starts making a funny noise, do you take photos of the engine to the garage and ask them to diagnose it?

You're showing an incredible amount of ingratitude.

[Katie Boundary]

If you're asking for help, you post the script and a sample. Basic procedure.

Normal procedure works for normal errors. This one is very unusual in that it cannot be reliably replicated, and any trial and error that we do right now won't yield very useful information. I've already explained this.

I'll be happy to delete this thread if you like.

I'd prefer to wait and see if anyone knows more.

[Sparktank]

I'd prefer to wait and see if anyone knows more.

I know exactly what the problem is.

[Groucho2004]

Normal procedure works for normal errors. This one is very unusual in that it cannot be reliably replicated, and any trial and error that we do right now won't yield very useful information.
Your logic is flawed. Just because you cannot reliably replicate the problem doesn't mean that it would be the same in another setup.
Empirical probability of an event is the ratio of the number of outcomes in which a specified event occurs to the total number of trials.

[Wilbert]

Closed pointless thread. Drop me a pm if you have a sample and script which shows the problem as manono requested, then i will reopen it.