MVTools and DeGrainMedian for interlaced source need advice

[yup]

Hi folk!
I write small script for interlaced source:
AviSource("sel.avi",pixel_type="YUY2")
AssumeTFF()
source=SeparateFields() #
bv1 = source.MVAnalyse(isb = true, truemotion=true, delta = 1, idx = 1, overlap=4, dct=1)
bv2 = source.MVAnalyse(isb = true, truemotion=true, delta = 2, idx = 1, overlap=4, dct=1)
fv1 = source.MVAnalyse(isb = false, truemotion=true, delta = 1, idx = 1, overlap=4, dct=1)
fv2 = source.MVAnalyse(isb = false, truemotion=true, delta = 2, idx = 1, overlap=4, dct=1)
bc1 =source.MVFlow(bv1, idx=1, thSCD1=1000)
bc2 =source.MVFlow(bv2, idx=1, thSCD1=1000)
fc1 = source.MVFlow(fv1, idx=1, thSCD1=1000)
fc2 = source.MVFlow(fv2, idx=1, thSCD1=1000)
interleave(fc2,fc1, source, bc1,bc2)
DeGrainMedian(limitY=9,limitUV=13,mode=1, norow=true)
SelectEvery(5,2)
Weave()

Please advise this script valid? Many script use separately even and odd fields, but if use all fields for same time I can use more data (3 frame equal 5 fields), predicted fields will be close to central fields by time and motion. Only one problem MVTools could estimate motion between fields up and down, but I think MVTools can this things.
yup.

[Boulder]

The resulting fields are spatially unaligned which means there'll be motion detected everywhere - why make the motion estimation any more difficult than it already is? :)

Bobbing would be a better choice IMHO.

[yup]

Hi Boulder!
If I right understand I need bobed before my script and after script use
SeparateFields().SelectEvery(4,0,3).Weave()
Yes?
But I spent my memmory and processor time for bobbed frame. I try my script and get normal results for VHS capture to DVD conversion. I can not find artefacts. I think MVTools without problem search vrertical shift from fields to fields.
yup.

[Boulder]

For the first question, yes.

Check the vectors with MVShow. Post screenshots from the MVShow output and we'll see what it looks like.

[Didée]

The vertical bobbing basically *is* a problem, but it also depends on a lot of different aspects. One of them is the "sharpness" of the source - with a very sharp & detailed source, the problems might get bigger. With a VHS capture, OTOH, it might be less important.

Anyway: yup, your script in fact is overcomplicated. DegrainMedian considers only prev/current/next frames to denoise the current frame. You're also doing MV-search and -compensation of the prev2 and next2 frames ... that is superfluid, since DegrainMedian won't consider them. So you can just leave out the acc. parts, and just interleave(fc1,source,bc1)

[yup]

Boulder!
I need append to my script two row
vectors = MVAnalyse(isb = true)
source.MVShow(vectors)
before Weave(), and disable Weave?
I try look without up down jerk.
yup.

[yup]

Hi Didée!
May be for big time lag use MedianBlurT or advice other temporal median softener.
yup.

[Boulder]

You could use bv1, bv2, fv1 or fv2 instead of defining "source".

[yup]

Didée!
I use DeGrainMedian(limitY=9,limitUV=13,mode=1, norow=true) repeately two times for 5 frames, I also try MedianBlurT with time radius 2 and not very satisfy by comparing to DeGrainMedian especialy during scene changes.
yup.