Personal View site logo
Crusade on MJPEG hacks
  • 73 Replies sorted by
  • bubbabubba
    a lot of work........
  • cosimo_bullocosimo_bullo
    "So emotional. Who's attacking MJPEG or the hack? I am fucking sick and tired of you guys blowing stuff up into a shit storm. If you can't be rational then go somewhere else."

    Read your own posts, Andrew. You have been attaching, and in a most unpleasant way. If you're ready to get constructive, welcome back.
  • Ian_TIan_T
    Knee grows please!!!!!!

    ...in the words of the (in)famous Rodney-the-King..."Can't we all just get along?" :-)
  • AriaAria
    From my personal perspective it would seem that the AVCHD path has more life to it. I'd be more interested in seeing how far that it can be improved upon. It looks to me to be starting from a better quality position and that's very important to me.

    Perhaps from my novice eyes the MJPEG seems to be about a stop brighter to me than the AVCHD clips EOSHD provided. Overall tho I didn't like the noise as much on the MJPEG clips. I'd be very interested to hear more tricks that can be done to fool the AVCHD into providing more detail in low light.
  • Vitaliy_KiselevVitaliy_Kiselev
    > I'd be very interested to hear more tricks that can be done to fool the AVCHD into providing more detail in low light.

    We must understand AVCHD encoder better to do this.
  • Vincent_PereiraVincent_Pereira
    Re: whether or not the LPowell 1080P MJPEG patch is "upscaled 720P" or not- couldn't some tests be done with resolution charts to figure that out once and for all?

    Vincent
  • cosimo_bullocosimo_bullo
    @Vincent: Take a look at Matt's last comment.
  • SkeptikalSkeptikal
    "So emotional. Who's attacking MJPEG or the hack? I am fucking sick and tired of you guys blowing stuff up into a shit storm. If you can't be rational then go somewhere else."

    Hello...My name is Pot...and you are a black Kettle...


    This acrimonious debate is very distracting...and pointless. If everyone had the same opinion on everything it would be pretty boring indeed.

    I use AVCHD about 99 percent of the time. Along comes a new patch for a new camera (for me anyway) which greatly improves how MJPEG looks. Surely it's up to each user to decide on the advantages and disadvantages (as personally perceived) and use the codec/bitrate/camera he thinks best for the situation.

    For example, as far as I'm concerned AVCHD in the GH2 is currently unusable FOR WHAT I DO because of the start-up glitch. 100mbps MJPEG (and particularly 30p) is very usable FOR WHAT I DO ...end of fucking drama...FOR ME.
  • EOSHDEOSHD
    It seems with the latest PTools that MJPEG quality is better than before especially in 4:3 mode. My main point was that MJPEG was not a low light killer. What I do find interesting is how Vitaliy has got the better quality out of 4:3. The first time I tried it with the first PTools release it looked like upscaled VGA and now when I put it to 1440x1080 for example it actually looks pretty damned close to 1440x1080.

    Yes the 1920x1080 detail is not quite 1080p but it is an improvement. Thanks Vitaliy.
  • SkeptikalSkeptikal
    To convert me from disliking/never using MJPEG to now actually using it for production/real world acquisition, I'd say the new 1080p MJPEG/GH2 is a very significant improvement. Well done all those responsible.
  • stoneystoney
    Hopefully we see improvements in mjpeg anamorphic setting, too.

    Who knows? One day 2544x1080 mjpeg in GH3vk or GH4vk... drooling...
  • Vincent_PereiraVincent_Pereira
    cosimo_bullo: Matt says he sees more details than 720p, but "not much though", and to come to that conclusion he says he downscaled a SCREENSHOT from a 1080P MJPEG video to 720P then upscaled said screenshot to 1080P... I'm sorry, but what the heck does that prove?

    My question was and still is, why doesn't somebody on EITHER SIDE of this debate actually do a real test using a resolution chart shooting natively in both modes?

    Vincent
  • ehrehr
    "The first time I tried it with the first PTools release it looked like upscaled VGA and now when I put it to 1440x1080 for example it actually looks pretty damned close to 1440x1080."

    So has something changed/ been improved in 4:3 MJPEG mode from the 1st PTools to the current version? If so this is great news!

  • RoryBRoryB
    Because debating is more fun?
  • cosimo_bullocosimo_bullo
    @Vincent. We will have a chance to do just that (or close to that) this weekend. In production this week with cameras tied up. I could try to find the couple of A/B tests I did with AVCHD and MJPEG at 1920/1080 but you're right, a res test would be best.
  • lenuisiblelenuisible
    >"The first time I tried it with the first PTools release it looked like upscaled VGA and now when I put it to 1440x1080 for example it actually looks pretty damned close to 1440x1080."

    with that sort of resolution difference, it should be quite obvious to see if VGA is upscaled. Could someone post some snapshots from a 1440x1080 MJPEG video ?
  • kaekae
    @All

    Wow. Just walked into this room. Is it hot in here or what?
  • ishvarishvar
    It's about Avchdism against Mjpegism and their repective video worshippers & -deities...
    Vitaliy spoke already about a crusade!
  • %5BDeleted+User%5D[Deleted User]
    I tend to agree with EOSHD's earlier points. I haven't touched the MJPEG video modes since I bought my GH2. It's largely a "toy" codec that finds its way into low-performance electronics that wanted to add "hd video capture" as a bullet point on the package. The GH2 is a prosumer camera, so yeah it has some consumer-grade elements on there but it makes little sense to use them when there are superior AVCHD modes available.

    If you want to send mom a video of your pet turtle riding a unicycle...MJPEG all the way....but did you really spend $2K+ on the GH2, lenses and support equipment just to record glorified webcam-quality video? If you are shooting video and care for quality, sticking with any of the AVCHD video modes would serve you better than the tweaked MJPEG modes. I should point out that AVCHD at AQ2 with 66 or 132m bitrates does well in low light motion and shadows...and that is something that has been tested by other members on here already.
  • RamboRambo
    @WeBcYtE>If you are shooting video and care for quality, sticking with any of the AVCHD video modes would serve you better than the tweaked MJPEG modes.

    Not so the GH1:-)
  • FrameFrame

    Sorry, no resolution chart but my CD-Collection. Stills were exported using mpegstreamclip. In the AVCHD Stills, Gamma was raised for better comparison. I guess, if I used the soft SLR-Magic 0,95 instead of the 14-140 lens kit for the "sharpness test", the difference in sharpness would be unnoticable. Not noticable in the stills but more noticable in the moving image: Lesser "mudding" in dark areas (I hope it is the right term) and lesser banding with MJPEG, I guess because of more noise? A sharper MJPEG with spanning (recording stops after 2:47mins/1.99gb on Sandisk SDXC Extreme Pro/64gb) and with better audio (seems to be restricted to PCM with 16kHz sampling frequency) MJPEG would be heaven to my eye. All the best, Frame

    SharpnessAVCHD.jpg
    1920 x 1080 - 2M
    Sharpness MotionJpeg.jpg
    1920 x 1080 - 1M
    Noktor_MJPEG_Flowmotion-1.jpg
    1920 x 1080 - 877K
    NoktorAVCHDQuantum100.jpg
    1920 x 1080 - 834K
  • vicxvicx
    According to cbrandin's recent investigations, the AVCHD encoder's highest quality quantization factors used for shadow detail are around 18-20.

    Is this information still accurate? Is there any thing that limits the IQ that can be achieved by low luminance regions (putting noise aside).

  • cbrandincbrandin

    That's not quite right. Depending on matrices, the effective QP can be much lower (lower is better). Going lower than 5 is a waste (because that's for 10-bit, not 8-bit). The new "soft" matrices produce an Effective QP of 5 (for DC value) in many cases, which is ideal, when the PTool quantization parameter is set to 18. Other settings can achieve this too. You can determine the minimum effective QP value by using the JM-SP decoder in StreamParser. The DC value is the lowest effective QP value for each frame. Basically, the quantization parameters by themselves are meaningless because they are processed with matrices to get the final, effective QP - which is what is important.

Add a Comment