Dots! Oh god, dots!!

Garylisk · Post by **Garylisk** » Thu Dec 28, 2006 1:57 am

It's not a problem with an AMV, but a video problem in general. I trust the AMV community to be quality whores and know how to handle crazy things like this, so I present you with...

EXHIBIT A

Circled below are the problem areas!

Notice the horrid dots! They move in the video, too. They seem to happen when lines are near one another, and Washu has the problem the most out of any aspect of the file.

The images you see above are taken without any filtration - that is the quality of the original video file. (A raw episode 1 of "Sasami - Mahou Shoujo Club")

Do you think it is even possible to somehow right this horrible wrong?

Kariudo · Post by **Kariudo** » Thu Dec 28, 2006 2:16 am

yes it is.
this is a well-established problem known as dot crawl.
there are a few filters that can take care of this in AviSynth.

<a href="http://avisynth2.cvs.sourceforge.net/*c ... ">Tcomb</a> (can be used for de-rainbowing and dot-crawl removal...part of Tritical's TVITC package)
<a href="http://avisynth2.cvs.sourceforge.net/*c ... ">undot</a>
dedot
(IIRC all of these filters need to be used before de-interlacting/ivtc-ing)

I have pictures of results I got from the R1 release of FMP Fumoffu using Tcomb in this thread

Garylisk · Post by **Garylisk** » Thu Dec 28, 2006 2:23 am

Apparently the dots only occur when there is motion.
Observe exhibits B through F

Absence of dots

Dots appear

Dots still apparent

Dots less apparent

Dots dissapear

Garylisk · Post by **Garylisk** » Thu Dec 28, 2006 2:24 am

Kariudo wrote:yes it is.
this is a well-established problem known as dot crawl.
there are a few filters that can take care of this in AviSynth.

<a href="http://avisynth2.cvs.sourceforge.net/*c ... ">Tcomb</a> (can be used for de-rainbowing and dot-crawl removal...part of Tritical's TVITC package)
<a href="http://avisynth2.cvs.sourceforge.net/*c ... ">undot</a>
dedot
(IIRC all of these filters need to be used before de-interlacting/ivtc-ing)

I have pictures of results I got from the R1 release of FMP Fumoffu using Tcomb in this thread

The source is progressive.

Kariudo · Post by **Kariudo** » Thu Dec 28, 2006 2:38 am

bummer...don't know of anything that can help in that situation.
maybe you can somehow interlace your footage, then do the dot crawl removal then de-interlace?

there might be something out there for vdub...maybe

Garylisk · Post by **Garylisk** » Thu Dec 28, 2006 2:47 am

Kariudo wrote:bummer...don't know of anything that can help in that situation.
maybe you can somehow interlace your footage, then do the dot crawl removal then de-interlace?

there might be something out there for vdub...maybe

It's doubtful, but I've kept trying things for a while... I don't see a solution since the footage is progressive.

Willen · Post by **Willen** » Thu Dec 28, 2006 6:43 am

Maybe Guava Comb? But if it's in the source and it's already progressive, that's gonna be tough. Someone messed up the broadcast or capture.

Scintilla · Post by **Scintilla** » Thu Dec 28, 2006 7:32 am

Shouldn't DeDot, etc. still work on progressive source? They always say to use before deinterlacing or IVTC in case the dot crawl was introduced after the telecine process, but if the source is progressive to begin with, then you don't have to worry about that.

If it helps, I have an English readme file (I did not translate it, somebody else did) for DeDot available here:
http://www.aquilinestudios.org/scripts/

Zarxrax · Post by **Zarxrax** » Thu Dec 28, 2006 12:04 pm

If it only occurs during motion, maybe the source has already been filtered? AFAIK, these dot crawl filters all work on a temporal-basis, so they cant handle motion areas.

Garylisk · Post by **Garylisk** » Thu Dec 28, 2006 2:13 pm

Zarxrax wrote:If it only occurs during motion, maybe the source has already been filtered? AFAIK, these dot crawl filters all work on a temporal-basis, so they cant handle motion areas.

Yep, I had to go out and find a different capture.