Personal View site logo
Make sure to join PV on Telegram or Facebook! Perfect to keep up with community on your smartphone.
Please, support PV!
It allows to keep PV going, with more focus towards AI, but keeping be one of the few truly independent places.
Interlacing artifacts in new 25p mode
  • Hi guys,

    I had a shoot of a band yesterday, and i wanted to share this picture with you.
    It does seem to have some interlacing artifacts at the top of the guitar.
    what do you think?
    Image3.png
    1920 x 1080 - 2M
  • 23 Replies sorted by
  • i think you'd see artifacts all over the frame if there was a problem. I suspect these are motion artifacts as opposed to interlacing issues.
  • but do you see clear lines there? do you mean compression artifacts by "motion artifacts"?
    If so, i think those are always blocky, correct me if im wrong.

    here is a closeup of the supposed interlacing artifacts
    blowup.jpg
    650 x 730 - 79K
  • What editor do you use? How did you import the footage?
  • i used sony vegas , set the footage to progressive and switched deinterlacing off. it could indeed be caused by the NLE. I just can't find a reason..
  • Also disable smart rerender in clip proppertys
  • This is compression noise. Don't worry about it. Remember that the camera is compressing interlaced video, which you are then weaving back to progressive.

    Mozes, do you mean smart resampling? That only matters when you are changing frame rates.
  • yhea sorry about that.
  • @balazer: ah, that could be it. so there is actually a downside to PsF. But Im hoping we will be able to eliminate this with the bitrate! :)
  • Guess the team will trow that much bandwith onto poor interlaced encoder that it will just give up and render proper images.. ;)
  • @ Balazer: The camera is not compressing any interlaced video. Psf is just a way of writing/reading/transporting full progressive stuff. Instead of going line by line you do it sequentialy, Odd lines first, even after. There s not any interlacing in that picture. These are 420 artifacts.

  • @Rafa

    You can have kind of interlacing artefacts, as compression is really same in interlacing case. They won't be the same, of course. But in case of different compression and deblocking for two half parts you can, rarely see odd and even lines.

  • @Vitaliy I agree with you. I think that's produced on conversion. What its come to my head is that the footage has been treated as interlaced. In that case the chroma of each original 2x2 Progressive block (made of correlative lines: 1-2, 3-4,..) would end up split (1-3, 2-4,..). That added to any 422 codec conversion (4 pixels in a row) could accentuate the effect and even more the high JPEG compression (horizontal) of the small pic. Would be interesting to know fatpig's workflow.

  • @Rafa

    Again.

    Here is how it works.

    25p footage -> 50i interlaced encoder -> etc

    Yep. This is how it works, Because I am looking at this right now. Encoder don't know anything about 25p or 30p modes.

  • What goes out at 25p via HDMI is definitely PsF, no interlacing to be seen there. So, if you want to avoid the interlacing encoder, get an external recorder like the Ninja.

    With the future hack of 1.1 you'll probably need to do severe pixel peeping to see a difference, though. I'm not sure it's worth the money for most users of a hacked GH2.

  • There's no interlacing in that picture. Look at the high speed motion of one of the string tuning things, it is smooth. Not combed.

  • @Vitaly, I don't catch it. If the CMOS can't be scanned at p25/30, I see only two options of getting p25/30: - i50/60 and de-interlacing (not a quality option). - p50/60 and dropping half of the frames (easy to do in real time). From p50/6O would be just a matter of writing down half of the frames to get p25/30 or half of the lines to get i50/60. If the CMOS is scanned at p25/30 and the signal is processed as i50/60, then is the processor who is introducing those interlaced like artifacts.

    @nomad, No really. I tried today recording HDMI on the NANO. Camera set to p25 and I'm getting fully interlaced stuff. No psf.

  • If the CMOS can't be scanned at p25/30,

    I already told you how it works. No need to such bunch of if's.

  • @Rafa

    Are you sure? Current firmware version from Panny, camera set to PAL, recording to 25p, recorded with an external recorder, telling the NLE it's progressive? I can see no interlacing then.

    Are you seeing any difference in our testing (apart from my use of a Ninja) ?

  • @Rafa, I think Vitaliy covered it pretty well, but just to be absolutely clear, this is what's happening in the camera in PsF mode: The sensor is capturing a sequence of progressive frames. The camera then splits those frames into fields, and feeds an interlaced signal to the encoder. The encoder then encodes that interlaced signal and produces an interlaced output sequence. Of course you and I know that the underlying video is progressive, and the original progressive frames can be reconstructed very simply by weaving the fields back into frames. But the encoder is entirely unaware of that. The artifacts in question are a result of the encoder operating in interlaced mode on an interlaced signal, with a bit rate inadequate to represent the original picture perfectly. The artifacts have nothing at all to do with 4:2:0 chroma sub-sampling.

  • @balazer: I got it now. What misleads me was using the term "encoder" instead of "video processor" (the encoding happens in the end writing the signal as AVCHD or MPEJ). Sorry if I ask too much. Is the only way to understand things. @nomad: I tried yesterday with the camera set to HBR and the HDMI set to AUTO. However I'm very new to the camera. I'll go through all has been posted in the forum about HDMI and repeat the tests.

  • Make sure you are at HBR and the HDMI is at PAL. If the HDMI is at NTSC, you'll get a very strange field cadence.

  • @fatpig @balazer @nomad , I've been making some tests shooting at HBR p25 with high ISO (PANASONIC 25.f1.4), and I find those artifacts everywhere. I've used 5DtoRGB, direct import to FC and Log&Transfer with the same results. The camera must be making some kind of crappy de-interlacing on the fly. The footage is unusable. In short: HBR is a shit. Is the same issue related in this tread: http://www.personal-view.com/talks/discussion/693/gh2-showing-fixed-darker-bandstripe-20-pixels-high-across-footage/p2

  • Can you post a full frame with fast horizontal motion?