PAL DV (interlaced) compression has movement stutter?
 

Hi,

I am currently trying to compress a PAL DV avi video (interlaced 25 fps) to mpeg2 format (CQ 100, max bitrate 3000) and I am getting stuttering on fast movements. Everything else looks very good.

I am using TMPGenc, the field order is set correctly and interlace mode is set for both input and target.

I am using the KVCD interlace matrix.

Im currently trying to deinterlace the video before compression and will post the result when finshed.

Thanks all.

If you do a resize in your script be carrefull to use a interlaced oriented resizer also !

Note : no need to go to CQ=100 as CQ=90 will give the same quality if a lot less bytes.

Re: PAL DV (interlaced) compression = movement stutter
 

Bottom field first? I know NTSC DV sources are BFF, and assume that PAL ones are too. But I could be wrong...

Yes, it is bff.
Also I do no resize.

After deinterlace (DGBob(0,0)) the movement is alot better.

I found a thread on videohelp.com which I believe is the same issue:
http://www.videohelp.com/forum/viewtopic.php?t=213662

Last post:

I'm not sure what the difference between frame-based interlaced and field-based interlaced is (at least with respect to TMPGEnc vs. Canopus Procoder), but all of the interlaced material I've encoded with TMPGEnc looks alright to me (even in high-motion scenes). I'll check the most recent DV I did, though...

I have not contibuted much to the subject anymore because it surpasses by far what I know.

But what I have been reading is that not every interlaced source is the same and that DV is special and will produce poor results on deinterlacing.

So, this does not exlude some interlace sources from being treated OK by TMPGenc on compression and others not.

The guys that wrote that probably forgot that the filed order is not the same as it used to be. If you don't take care to that, the result is necessary bad.

I don't see why : there is ONE way to interlace. As soon as you have a script that gives a correct result, it should work for any source of the same kind.

As u use a high AVG and TmpgEnc's Q of 90 .....

... why dont you keep it interlaced?? As you dont suffer from a reduced outputspace finally??

Did you try to change the fieldorder??

Separatefields()
Trim(1,0)

Theoretically this changes the bff to tff (no swap fileds!) ... if its really needed.

A truly interlaced DV Video is fieldbased. As on motion scenes every field gots its unique motion, ... thats why interlaced videos do look smoother and why there are combs.

In Avisynth the word "Fieldbased" is a bit hadled confuding as in avisynth "fieldbased" means that the fields have been separarted (full fieldrate by 1/2 height).
If your "fields" are in a weaved state (as your video comes out of the camera) then Avisynth "sees" even that as "Framebased".

So In avisynth "Framebased" = weaved fields
"Fieldbased" = separated fields

Okay, that makes sense... What about TMPGEnc? Does it do what I'd expect and encode each field separately when interlaced output is selected?

I have done some tests with Canopus Procoder Express and the results were very good, no stuttering movements at all and the image is very clear.

So it is an issue with Tmpgenc with interlaced PAL DV material.

This seems to have been confirmed by some people on different message boards. The overall advice is always to de-interlace (which I dont want to) or switch the encoder.

Sadly, this means Tmpgenc is unusable for me. All those hours will not really be lost, but now that I was used to it... Oh well.