Personal View site logo
GH1 - Real 1920x1080 MJPEG possible?
  • enimenim
    Can the MJPEG frame size @ 30 FPS be increased? Is 1920x1080 achievable? I'm looking at "MJPEG Size Adjustments" in PTool v.3.61d, and that's what it says. But does it just do upscaling, which should probably be left to editing, or are the individual frames shot by the camera actually that size?
    I'd appreciate any input you guys might have. Have been reading lots of discussions here with great interest.
    Thanks!
  • 16 Replies sorted by
  • fatpigfatpig
    at least until now its always been upscaling. ive tried the new 100mbps mjpeg 1920 x1080 patch, its upscaled. sadly.
  • LPowellLPowell
    The 100Mbps Low Light MJPEG Patch works only on the GH2. For the GH1, the largest frame size I was able to produce was 1536x864. You can find that patch here:

    http://www.personal-view.com./talks/discussion/334/75mbps-gh1-peak-reliability-patch-v2-update#Item_11

    While numerous posters have speculated that patched MJPEG frame sizes are actually upscaled from a 1280x720 frame, I have yet to see any tangible proof of this claim. My recommendation is to try it yourself and see what you think.
  • EOSHDEOSHD
    Is it yet possible to change the frame rate to 24p? I hate 30p with a passion.
  • cbrandincbrandin
    It's amazing - the psychological difference between 24fps and 30fps. I suppose the human eye resolves at just below 30fps and above 24fps.

    Chris
  • enimenim
    @LPowell

    Michael Pappas (PappasArts) seems to have achieved 1440x1080 with his MJPEG 70mbits+ settings, which would be a bit bigger than 1536x864 albeit anamorphic. Don't you need special lenses for that sort of thing?
  • LPowellLPowell
    On the GH1, GF1, and G2, I was never able to find 1920x1080 settings for MJPEG that would record reliably in all circumstances. Narrower frame sizes, such as 1440x1080, produce distorted videos that appear stretched horizontally. Unfortunately, this is the opposite type of distortion than you want to use with anamorphic lenses. For that purpose, I found anamorphic 1.33X 1920x810 and 1.5x 1920x720 frame sizes to be ideal.

    The 30p -> 24p MJPEG patch has never worked properly. It records at 30p and simply discards every fifth frame, producing choppy motion. As far as I can tell, 30p is hard-wired into the MJPEG encoder.
  • Vitaliy_KiselevVitaliy_Kiselev
    >It records at 30p and simply discards every fifth frame, producing choppy motion. As far as I can tell, 30p is hard-wired into the MJPEG encoder.

    Nope. 30p is not hard-wired.
    It is video feed from sensor that is 30p.
    MJPEG encoder work properly using 24p.
  • LPowellLPowell
    Right, but my impression was that the GH1's MJPEG encoder was hard-wired to the 30p sensor mode. Can this now be patched to 24p on the GH2?
  • Vitaliy_KiselevVitaliy_Kiselev
    @LPowell
    Encoder can be patched, same as in GH1.
    But it is video setup that not allows us to have smooth and good 24p.
    This think is fuckingly complicated with values moved to structures, then to other structures, then to other structures, then processed, then moved to other structures, etc. :-)
    Big fucking fun.
  • fatpigfatpig
    @lpowell:
    the first thing i saw when testing the patch on gh2 (1920 mjpeg) is that its not nearly as sharp as avchd 24p. Maybe I can do a test shot, im sure you did see that too? were there no test charts done with this to compare?

    @Vitaliy:
    So highest goal would be to combine gh2 sensor mode from 24p with mjpeg encoder, right?
  • RamboRambo
    Aren't the mpeg macro blocks larger than AVCHD, wouldn't this effect how we perceive resolution?
  • LPowellLPowell
    At low bitrates, the rectangular edges of the MJPEG macroblocks can become visible under magnification. At 90-100Mbps, the encoder appears to have enough bitrate to eliminate this problem.
  • enimenim
    So to sum up, we can't rule out that MJPEG delivers true 1920x1080, but from the looks of it that possibility seems unlikely.
  • mpgxsvcdmpgxsvcd
    MJPEG is always upscaled. Take a look at these screen shots and you will see.

    http://forums.dpreview.com/forums/readflat.asp?forum=1041&message=38912742

  • LPowellLPowell
    @mpgxsvcd - Thanks again for the link to your test chart shots. There are three GH2 samples I found quite interesting:

    GH200001 MJPEG 720p 0 IR_resize (Original) - MJPEG 720p upscaled to 1080p
    P1350105 GH2 MJPEG 1080p 0 IR (Original) - MJPEG 1080p
    GH200001 AVC 1080p 0 IR (Original) - AVCHD 1080p

    In most details, all three images look very similar, indicating that the GH2's optical resolution at 1080p isn't much higher than 720p to begin with. In addition, I spotted evidence of sensor sampling artifacts present in the upscaled 720p shot that are not present in the two 1080p shots.

    In the lower middle section of the test chart, there's a row of diagonal line patterns labeled 1 3 5 7 9 10 8 6 4 2. In the upscaled MJPEG 720p shot there is a clearly visible vertical thickening of the diagonal lines in the middle of the pattern labeled "7". This is a sensor sampling artifact that can also be discerned in patterns 9, 10, 8, and 6.

    If the MJPEG 1080p shot was in fact upscaled in-camera from a 720p frame, it should show the same type of sensor sampling artifacts as the 720p shot. In this case, the upscaled MJPEG 720p sampling artifacts in patterns 7, 9, 10, 8, and 6 are NOT present in the MJPEG 1080p shot.

    Under detailed examination, the 1 3 5 7 9 10 8 6 4 2 diagonal line patterns in the MJPEG 1080p shot exhibit the identical sensor sampling artifacts as the AVCHD 1080p shot. This indicates that both MJPEG and AVCHD 1080p shots were most likely downsampled from the same high-resolution source (the 16-megapixel image sensor).
  • enimenim
    LPowell - saving the rest of us a bunch of time yet again! :)
Add a Comment