How to fix Chroma Bleed with AviSynth?

[jagabo]

I actually thought the problem with the video was that it looks like there's too much red around the ropes - they look strange. What did you guys think the problem was?

Yes, there's to much smearing of the chroma channels. There's no fixing it.

[VideoFanatic]

Can the following be used on interlaced video without issues or would I need to convert to progressive with QTGMC then use the script then reinterlace?:

Overlay(GreyScale(last), last, y=-4, mode="chroma")
MergeChroma(awarpsharp2(depth=35))

[jagabo]

Overlay() is ok since you are moving vertically by 4 scan lines (mod 4 with YV12, mod 2 with YUY2). MergeChroma() is OK. But awarpsharp() shouldn't be used on interlaced video.

By the way, there's no need to use GreyScale() in the Overlay. You're replacing the chroma channels so there's no need to eliminate them first. Ie, all you need for that line is:

Overlay(last, last, y=-4, mode="chroma")

[VideoFanatic]

Thanks for the tip. The first lines in my script are like this:
McTemporalDenoise(settings="medium", interlaced=true)


AssumeTFF()
QTGMC(preset="Super Fast")
SeparateFields() SelectEvery(4,0,3) Weave()

I normally use QTGMC on interlaced footage then reinterlace. However if a script is only supposed to work on progressive footage yet I need to run the script before QTGMC then how would I do that? For example DeBlock_QED() needs to run before McTemporalDenoise doesn't it so what should I do?

[sanlyn]

Thanks for the tip. The first lines in my script are like this:
McTemporalDenoise(settings="medium", interlaced=true)


AssumeTFF()
QTGMC(preset="Super Fast")
SeparateFields() SelectEvery(4,0,3) Weave()

I normally use QTGMC on interlaced footage then reinterlace. However if a script is only supposed to work on progressive footage yet I need to run the script before QTGMC then how would I do that? For example DeBlock_QED() needs to run before McTemporalDenoise doesn't it so what should I do?

AssumeTFF()
QTGMC(preset="Super Fast") 
DeBlock_QED()
SeparateFields().SelectEvery(4,0,3).Weave()

[VideoFanatic]

I know I can do that but I was under the impression that DeBlock_QED() should be run before any other cleaning filters? If that's true I can't use Deblock_QED after QTMC because McTemporalDenoise is run before QTMC. Running McTemporalDenoise after QTGMC gives an FPS of under 1!

[juhok]

From the aWarpSharp2 README you can find that:

MergeChroma(awarpsharp2(depth=35))

is the same as:

aWarpSharp2(depth=35, chroma=6)

QTGMC(preset="Super Fast") is a bit overkill and destructive method to bob before a filter if your intent is to keep it interlaced. Correct DeBlock_QED() usage with interlaced source (you could've found this with 5 secs of google):
http://forum.doom9.org/showthread.php?p=934083#post934083

aWarpSharp2 for interlaced source (AFAIK):

SeparateFields()
aWarpSharp2(depth=35, chroma=6)
Weave()

[VideoFanatic]

Thanks for the info juhok. So basically to use a Deblocker on interlaced footage before any other filters I do this, is that correct:

SeparateFields() .PointResize(width,height)
  Deblock_qed() .AssumeFrameBased()
SeparateFields() .SelectEvery(4,0,3) .Weave()

I read this:

PointResize is the simplest resizer possible. It uses a Point Sampler or Nearest Neighbour algorithm, which usually results in a very blocky image. So in general this filter should only be used, if you intend to have inferior quality, or you need the clear pixel drawings. Useful for magnifying small areas for examination.

So is the red text going to do any harm to my video?

So basically if a filter only works in progressive but I need to use it before QTGMC then I should do the following (replacing the middle line with whatever filter I'm using), is that correct?

SeparateFields() 
aWarpSharp2(depth=35, chroma=6)
Weave()

][QUOTE=juhok;2259596]From the aWarpSharp2 README you can find that:

MergeChroma(awarpsharp2(depth=35))

is the same as:

aWarpSharp2(depth=35, chroma=6)

I don't understand the relevance. They both do the same thing don't they so does it matter which one I use?

[juhok]

In this specific case PointResize is good for what you're doing. Didée is the author of Deblock_QED() and knows his stuff.

SeparateFields() -> Filter -> Weave() does not work well in every case. You should avoid temporal filters which are not aware of the SeparateFields() situation.

For temporal filters you could use lossless bob, for example

Bob(0.0,1.0)
Filter
SeparateFields().SelectEvery(4,0,3).Weave()

Or if you like the slowness, run QTGMC in lossless mode.

From the aWarpSharp2 README you can find that:

MergeChroma(awarpsharp2(depth=35))

is the same as:

aWarpSharp2(depth=35, chroma=6)

I don't understand the relevance. They both do the same thing don't they so does it matter which one I use?

MergeChroma is redundant operation, a leftover from the time when aWarpSharp didn't have that option. Why complicate things? I ran AVSMeter on both versions of the script and speed difference was neglible, so there's no penalty either way.

[sanlyn]

I know I can do that but I was under the impression that DeBlock_QED() should be run before any other cleaning filters? If that's true I can't use Deblock_QED after QTMC because McTemporalDenoise is run before QTMC.

Maybe you're confusing deblockers with anti-alias. The latter should usually be run before deintelacing.

Running McTemporalDenoise after QTGMC gives an FPS of under 1!

The script doesn't run MCTD after running QTGMC. It ran MCTD first. All I did was copy your own script posted earlier and inserted DeBlock_QED in the most useful sequence.