Log in

View Full Version : Use decross before or after ivtc (source)?


HOLiC
14th February 2006, 11:28
I have been trying to find out more information about this filter, but there is almost no information out there!! However, i did managed to find the japanease site with this filter (not that i understand japanease). And it seems like it's used for getting rid of rainbows and something about "cross colors".. is this correct?

anyway, what I want to know is if it is okey to use this filter before you ivtc your source, like this:

mpeg2source()
decross()

or is this wrong? thank you~

Mug Funky
14th February 2006, 12:50
run it before IVTC. IVTC will mung the field order, which is good if you want progressive frames, but bad if you're using the actual order of fields to cancel stuff out. dedot is this way - it must be run before anything else as it needs the exact fields that were captured.

also, if decross is successful it'll reduce false matches in your IVTC, giving a better result there as well.

Didée
14th February 2006, 13:54
A link to decross, for the busy people?

Chainmax
14th February 2006, 16:05
Hey, it's on the same website as DeDot (http://nullinfo.s21.xrea.com/)! :) It's probably made by theJam too then, right?

niiyan
14th February 2006, 17:11
@Chainmax:

It's probably made by theJam too then, right?
It was made by minamina, referring to the crosscolor plugin for AviUtl (crosscolor.auf) (http://latch-up.hp.infoseek.co.jp/aviutl.html).

@all:

Here is a part of Japanese-English translation of DeCross.txt.

- Usage

LoadPlugin("DeCross.dll")
DeCross(clip, ThresholdY, Noise, Margin, Debug)

ThresholdY : Threshold for edge detection (0 to 255, default 30)
Noise : Considered as noise if more or equal than this value. (0 to 255, default 60)
Margin : The edge detection part is expanded right and left. (0 to 4, default 1)
Debug : Show the area the filter applied (default false)

Chainmax
15th February 2006, 13:55
I see, thanks. Could you please post the complete translation?

niiyan
15th February 2006, 16:08
Here:

DeCross.dll v0.0.02 Copyright(C) 2003, 2004 minamina

Avisynth Plugin - Cross Color Reduction (YUY2 and YV12 Only)


- Explanation

Cross Color reduction plugin for Avisynth.
It supports YUY2 and YV12.
Avisynth 2.5 only.
MMX required.

It was made by referring to the crosscolor filter for AviUtl in Shou's Page.
Thanks Shou.


- Usage

LoadPlugin("DeCross.dll")
DeCross(clip, ThresholdY, Noise, Margin, Debug)

ThresholdY : Threshold for edge detection (0 to 255, default 30)
Noise : Considered as noise if more or equal than this value. (0 to 255, default 60)
Margin : The edge detection part is expanded right and left. (0 to 4, default 1)
Debug : Show the area the filter applied (default false)


- Example

DeCross()
DeCross(28, 80, 2, false)
DeCross(ThresholdY = 40, Margin = 1)


- History

2004/02/01 v0.0.02 Fixed the problem that it couldn't be read by using LoadPluginEx + AviSynth 2.0.x.
2003/12/26 v0.0.01 Initial Release.

Chainmax
15th February 2006, 16:23
Thanks :). It doesn't say wether it can be applied before IVTCing/Deinterlacing though.

foxyshadis
15th February 2006, 21:39
It might be the same as bifrost, ie, use it depending on whether it was introduced before or after telecining. I don't know if it's similar enough though.

niiyan
16th February 2006, 17:08
The documentation of the AviUtl's crosscolor plugin says:

- Directions -

If you don't use it before deinterlacing, this filter has no effect.
It doesn't work, except when the width of the image is 640,
704 or 720.

I don't know whether that is true of DeCross.

Wilbert
16th February 2006, 21:13
You always need to apply such filters before any deinterlacing, because it uses the periodicity of the signal to get rid of the noise. Deinterlacing destroys that periodicity.

Chainmax
16th February 2006, 22:03
But do all these kind of filters support interlaced processing?

Mug Funky
18th February 2006, 15:17
it really depends on what assumptions are made about the artefacts.

a generic "spatial chroma nuke" de-rainbower would work equally on interlaced or progressive material, because it'd use only basic and brutal means of detection and elimination. some actually need an interlaced picture to work, though.

after all, rainbows/dot-crawls only occur with composite video, which is _always_ interlaced regardless of whether the actual signal contains an interlaced or progressive (or telecined) picture. so the crosstalk that comes from it will be per-field, and thankfully first and second fields cancel out (which is why it's actually quite hard to see rainbows on a TV fed with composite - your eyes work slower than the field rate, so it largely cancels out before it hits your brain).