Personal View site logo
Real MJPEG 1080p resolution.
  • Hi everyone,

    I'm not sure is this is the right way to ask a question of everyone on the issue of resolution regarding the Mjpeg codec but I didn't see a category anywhere that pushed this issue.

    The reason why I have a concern over this is that I noticed with the ex. Telescopic conversion (ex.tel) on that when shifting between 24H and 720p 60p the image automatically zooms forward a bit more. Since in ex-tel mode the camera is no longer downsampling off the entire 4k sensor it makes sense since this is the equivalent of a digital zoom. It uses a central 1080p sized piece of the sensor and then it zoomed in even further to get a smaller 720p sized piece of the sensor. If I switch it VGA mode it the image will zoom in just a tiny bit more. I can only assume it is at that point jumping into an even smaller 480p sized piece of the 4k sensor.

    This is all fine, what worries me is that if I switch back to 720p60p and am thus on the 720p sized piece of the sensor, and then switch to HD Mjpeg, which I have set for 2K (2048 by 1152) level of resolution, the image does NOT, repeat does not jump back out to at 1080p or 2K sized image from the sensor. It's exactly the same. This suggests, in ex-tel mode that as only a 720p sized piece of the sensor is used to film 720p60p mode, the exact same sized piece of the sensor is used to film HD MJPEG mode as well. In other words even if you reset the HD MJEG resolution away from its native 720p settings and push it up to 1080p or even 2K, it may still be only recording 720 level of resolution. In other words its possible the HD Mjpeg system merely upsamples its native 720p footage such that when we assess that footage in quicktime or VLC it will state that it is 1080p or 2K but in terms of literal optical quality we still only have 720 lines to play with.

    I've noticed with identical shots I've taken with HD MJPEG 2K and 1080p 2p AVCHD that the resolution was almost identical but I somehow felt there was more clarity off of the 1080p footage. Technically the 2K I had with 2048 by 1152 should have been a bit better, but it wasn't.

    Has anyone done a resolution test on the HD MJPEG, 720p and 1080p modes to confirm what kind of resolution we're actually getting off these different codecs?
  • 4 Replies sorted by
  • Let's hope this "if i use my anamorphic lens i lose resolution" trade-off will be a thing of the past when the mp4 arrives. Oh god i hope they'll include mp4 for the VGA mode. Provided Vitaliy finds it in him to hack the next firmware, again..
  • Even when the GH2 is in ETC mode, it still has to use its demosaic filter to produce each video frame. Although the sensor resolution in FHD ETC mode is 1920x1080, there aren't actually 1920 red, green, and blue photocells in each of the 1080 lines. There are only half as many green photocells and only a quarter as many red and blue cells in the RGBG Bayer array on the image sensor. In all cases, the GH2 must use interpolation to produce the red, green, and blue components of each of the 1920x1080 pixels in an FHD video frame.

    Although Panasonic has not revealed the technical details of its demosaic algorithm, it is clearly an algorithm capable of working with a wide variety of image sensor resolutions. For example, it works just as well with the GH1's 12Mpixel sensor as the GH2's 16Mpixel sensor, producing the same size 1080p video frames from both. When the GH2 is put into ETC mode, it is effectively using an image sensor of about 1-2Mpixels instead.

    In addition to its ability to work with varying sensor sizes, the GH2's demosaic filter can also produce a variety of interpolated frame sizes, i.e. 1920x1080, 1280x720, 640x480, etc. When you select a video mode, the GH2 sets the video frame size to dimensions set in its firmware. In the case of MJPEG HD and VGA modes, Vitaliy has given us the ability to patch these firmware dimensions to any frame size we prefer.

    When you put the GH2 in ETC mode, the camera reduces the effective size of the image sensor by a digital zoom factor that's preset individually for each video mode. The selected video frame size, however, remains the same, and the demosaic filter interpolates the scanned section of the image sensor into that video mode's frame size.

    In FHD video mode, ETC has a 2.6X digital zoom factor, using a 2Mpixel center section of the 16Mpixel sensor. In HD video mode, ETC has a 3.9X digital zoom factor, using a 1Mpixel center section of the 16Mpixel sensor. Thus HD video has half the image sensor resolution as FHD video in ETC mode.

    With ETC off, however, both FHD and HD video mode use the full 16Mpixel image sensor to produce their video frames. Thus HD video has the same image sensor resolution as FHD video in non-ETC mode.
  • Would it be possible to raise / change the sensor size the cam uses in ETC mode? Or is it locked to 1:1 ? We could make much better use of Cmount lenses this way.