MVDegrain2 for interlaced source

[yup]

Hi folk!
Recently i try MVTools V1.6.2 by Fizick and find two function for denoise MVDegrain1 and MVDegrain2. When I use MVDegrain1 for interlaced source from examples I get good result, in CCE emcoding Q decrease from 7.5 to 5 at standardt set. I try write script for MVDegrain2 and interlaced source:
source=AviSource("Digit.avi")
fields=source.AssumeBFF().SeparateFields() # or AssumeBFF
backward_vec1 = fields.MVAnalyse(isb = true, delta = 2, pel = 2, overlap=4, sharp=2, idx = 1)
forward_vec1 = fields.MVAnalyse(isb = false, delta = 2, pel = 2, overlap=4, sharp=2, idx = 1)
backward_vec2 = fields.MVAnalyse(isb = true, delta = 4, pel = 2, overlap=4, sharp=2, idx = 1)
forward_vec2 = fields.MVAnalyse(isb = false, delta = 4, pel = 2, overlap=4, sharp=2, idx = 1)
fields.MVDegrain2(backward_vec1,forward_vec1,backward_vec2,forward_vec2,thSAD=400,idx=1)
Weave()
Question I right use MVDegrain2 for interlaced source? At some frame I get strange result with single points at luma border. May be someone advise how improve script use MVDegrain1 or MVDegrain2 for interlaced source, both script use bottom field and top field separately, but realy fields coupled. How can use information from one field to other during for precise estimation motion vectors?
With kind regards yup.

[Fizick]

why new thread? do yo ever try search with words:
mvdegrain2 interlaced
?

[yup]

Hi Fizick!
I try search and find only 14 threads, in all use fields separately without any coupling. If use information from other fields we will have more information, in theory 2 times and can decrease dispersion for motion vector estimation and also increase number point on time scale for averaging that give decrease noise. But I think this feature require work on plugin leven and can not be realised on script level.
yup.

[Fizick]

my general rule is not discuss mvtools-related questions in non-mvtools thread, and similar for other.
Bye.

[juhok]

Based on suggestion by Boulder (PAL material)

fields = LeakKernelBob(order=1,threshold=0)

backward_vec2 = fields.MVAnalyse(isb = true, delta = 2, pel = 2, blksize = 16, overlap=8, sharp=1, idx = 1)
backward_vec1 = fields.MVAnalyse(isb = true, delta = 1, pel = 2, blksize = 16, overlap=8, sharp=1, idx = 1)
forward_vec1 = fields.MVAnalyse(isb = false, delta = 1, pel = 2, blksize = 16, overlap=8, sharp=1, idx = 1)
forward_vec2 = fields.MVAnalyse(isb = false, delta = 2, pel = 2, blksize = 16, overlap=8, sharp=1, idx = 1)

fields.MVDegrain2(backward_vec1,forward_vec1,backward_vec2,forward_vec2,thSAD=200,idx=1, plane=0)

AssumeBFF().SeparateFields().Selectevery(4,0,3).Weave()

order=0 and TFF for TFF

I take no responsibility! :)

[yup]

Hi juhok!
i think use MVDegrain2 for bobed clip same like MVDegrain1 to source clip. I try my script (see begin thread) he work, I get decrease quant scale almost 2 times.
yup.

[Serbianboss]

I think that is safe to use just with Separatefields/weave because with bobing its very very slow

See this thread:
http://forum.doom9.org/showthread.php?p=912823#post912823