FixBlendIVTC

Latest revision as of 04:13, 11 November 2017

[edit] Description

FixBlendIVTC is a function to restore blend-deinterlaced telecined footage with double blends and 50:50 blend weights.
Note: Decimate (from Decomb) is the only function guaranteed not to mess up this filter, unless you follow the filter with RequestLinear from TIVTC.

[edit] Requirements

• AviSynth 2.5.8 or later
• Supported color formats: YV12

[edit] Required Plugins

Latest versions of the following filters are recommended unless stated otherwise.

• Average
• RemoveGrain - part of the RemoveGrain v1.0b package.
• RgTools - recommended drop-in replacement for RemoveGrain!
• MaskTools2

[edit] Syntax and Parameters

FixBlendIVTC (clip input, "post", float "bthresh", "mthresh", bool "sbd", clip "dclip")

input clip =

Input clip.

post int (0-6, default 2)

Postprocessing strength: Post=0 is the fastest, post=5 is the best for bad quality and the slowest postprocessing mode.

• 0 - No postprocessing
• 1 - 0 + chromablur
• 2 - Diffmasked pp (sharp and good for hq sources) - default
• 3 - 2 + chromablur
• 4 - Blurmasked restoring (more soften than 2, but sometimes less artefacts)
• 5 - Combines post 2 and 4 but without chroma.
• 6 - 5 + extra chroma (slowest).

If these postprocessing modes aren't strong or individual enough for you you can also use your own favorite filter on the restored frames. Therefore set a string with the filter(s) of your choice. Use pp0 till pp6 for the postprocessing mode and add your filter like this:

post="pp4.blur(1)"

If you want to use a filter with input-strings use three quotation marks like in the second example.

bthresh float (0.0-2.0, default 0.1)

The blend-threshold can be used to make the blend-detection less aggressive if there are some wrong detections.

mthresh float (0.0-1.0, default 0.1)

The motion threshold for the function. If the frame-difference is smaller than mthresh the function don't trust the blend-detection and does nothing. This helps decimating and can avoid side-effects. Sbd enables already some kind of mthresholding, but the 12 fps detection works with really low thresholds.

sbd bool = false

Single-Blend-Detection, it's a bool parameter you should set true for 12fps sources (anime/cartoons).

dclip clip = input

This clip is used for the blend-detection and won´t be returned, so that you can preprocess for blend-detection without affecting the output. A light blurring or temporal stabilizing can be really useful sometimes, especially in a source with many artifacts, but normally this feature shouldn't be used.

Note: The blend-detection-differencing is also used for post 4 and 5. So a too strong preprocessing could maybe have a bad effect on the postprocessing, but I wasn't able too produce such cases till now. Indeed sometimes this seems to be just better for the postprocessing.

[edit] Examples

AviSource("blended.avi")
FixBlendIVTC(post=3)
Decimate(cycle=5,quality=0)
LimitedSharpenFaster()

FixBlendIVTC(post="""pp2.Deen("a2d")""")
TDecimate(rate=23.976,mode=7)

[edit] Changelog

Foxyshadis

v.1 - v.3:
Required manual offsets. v.1 had broken reversal.
v.4:
Conditional, effective but very slow.

MOmonster

v.5:
Much faster, better detection & single-blend detection.
v.6:
Changed parameters again, better motion mask/influence.
v.8:
Sped up significantly, also more accurate single-blend.
v.9:
Modified for less artifacts, more post-processing choices, bthresh.

[edit] External Links

• Doom9 Forum - FixBlendIVTC discussion.

Back to External Filters ←