CCE 2.66 jittering problems

[cherry creek]

I am testing the CCE 2.66, but the results so far are quite poor regarding the video quality. I am trying to produce PAL SVCD with max output 2350 kbit/s, and am resizing and frameserving the clip with the avisynth.
The problem is that in fast movements there seems to be two horizontal "jittering zones", each one third of the screen up/down from the bottom/top. These are really distracting and the result is unacceptable. Does someone have experiences in this?

[Arky]

Sounds like incorrect field-order to me...


Arky ;o)

[cherry creek]

The field order is correct, DV - lower field first. I tried also the mainconcept encoder and the result is much better without jitter, but there are some artifacts there.
I would suspect that this is somehow related to mpeg motion vector settings. Does anyone know whether those can be adjusted in CCE? MC allows those to be changed, and I could try out the same settings with the CCE.

[jtor]

I suggest you re-encode your AVI in CCE, but this time, click the UPPER FIELD checkbox.

Yes, I'm aware of DV as lower field first. That's the setting that I've been using in TMPGENC. But, I also noticed this jitters in MPEG2 encoded by CCE (I'm not sure if we're looking at identical problems though), and I was able to get around it by doing UPPER FIELD first. They were probably looking at the frames upside down, when they made this upper field/lower field definitions (hehehe).

hope this helps.

[King Dick]

I use CCE 2.66 for all my encoding, and "jtor" is correct - upper field first box needs to be checked.

[cherry creek]

Thank you for the tips. Unfortunately no proceeds with any combination of upper field flags (also tried pulldown to change the field order after the CCE). If I change the field order, most of the video then jitters.

As the quality of the end result is ok for most parts, except for couple fast moving sections, I guess this relates to the encoding of the moving parts. Weird thing is that the MainConcept encoder handles these parts perfectly.

[celtiber2]

I've had similar problems when capturing/encoding old Hi8 tapes via my DV cam. So I don't know if the solution is applicable but in the hope it helps here goes...

My problems were when I frameserved from Premiere into CCE 2.5 using videoserver, I had 4 or 5 five bands of shuddering when there was movement. Like you I tried the different settings for the Upper Field First flag with no success.

Instead I ended up frameserving into TMPGenc 2.56plus and deinterlacing the video there (not in Premiere), but still producing an interlaced M2V file(for DVD/TV viewing). I used the Even-Odd Field (field) filter as recommended on other threads.

I don't really understand why this worked but think it has something to do with the different number of lines in the original tape compared with the DV conversion in the camera.

[cherry creek]

Sounds interesting, and applicable to my case. The sections with the problem were, indeed, shot with old svhs camera, and then transferred to DV. I will try out your solution during the weekend - still too tired after last nights fireworks and filming...

[cherry creek]

This is getting interesting: It just struck my mind when looking at the avisynth feed with the Windows media player that my source is DV Type-1, which forces me to use directshowsource in the avisynth script. From another tryout I know that the Windows DirectShow does not properly support interlaced sources i.e. you might have significant problems with that (I had the exact same problem with Muvee video creator when saving in DV mode) .
Microsoft are supposed to fix that to the DirectShow9, which is due some unannounced time this spring. This would also explain why main concept does not have any problems with my source file. Does anyone have any good suggestions how to overcome this? How could I save the file in DV Type-2 mode so that I could use avisource in the avisynth script?

[cherry creek]

DirectX9 was actually released Dec 20th. Here's some more info (copied from MSDN site):

http://msdn.microsoft.com/library/default.asp?url=/library/en-us/directx9_c/directx/ht...oderfilter.asp

- clip -

Interlacing: Earlier versions of the decoder always deinterlace the video. As of DirectX 9.0, the DV Video Decoder can preserve the interlacing. This enables the interlaced video to be deinterlaced by the Video Mixing Renderer (VMR), for improved rendering quality.

- clip -

This definately has an impact if the avisynth is using MS DV Video Decoder for frameserving.

[cherry creek]

FINALLY - CASE SOLVED:

Three different reasons:

1. as I posted earlier, MS DirectShow earlier deinterlaced everything and did not maintain the interlacing field information -> corrected by installing the new DirectX9 that was released Dec 20th.

2. my own avisynth BiLinearResize statement. I did have BilinearResize(480,576,8,0,704,570) to avoid flickering edges, but this did cause the shuddering zones due to the resolution change -> corrected by using simple BilinearResize(480,576).

3. now when the field order information was maintained correctly - guess what -> I did have to check the upper field first flag to avoid flickering video.

=> When I fixed all these I am able to have appropriate results with CCE.

I did all this to figure out the quality difference between the CCE and MainConcept, and have to admit that I did this for nothing. I am more satisfied with the MC results when I enable large motion search area. I do not mind long encoding times as my clips are always only a few minutes long. My decision on the software is then quite obvious considering the tenfold price difference. But anyway, I hope you can benefit of these experiments.

[troyvcd1]

I think there needs to be some post processing with pulldown after you encode with cce. Am I right? Does anyone know the settings in Pulldown to do this???