Archive for the ‘artefacts’ Category

RE:Vision’s FieldsKit ReInterlacer

Friday, May 16th, 2014

In Summary:

Purpose of FieldsKit ReInterlacer:

  • Transforms progressive video (e.g. HDp25 frames/sec) into spatio-temporal interlaced video (e.g. SDi50 fields/sec).  It achieves this by estimating the fields that would have been shot (had the original video itself been shot as interlaced) between each frame of the progressive video, via a process of motion estimation.
    •  Most NLEs do not use this “perfectionist” method, instead they at best simply combine (ghost-blur) successive frames, with no compensation for time/motion.
    • On an interlaced display, such as an old analog TV or projector,
      • The “NLE-simple” approach may lead to dynamic (changing e.g. moving) scenes and objects appearing flickery.
      • The “perfectionist” approach will instead typically avoid such flicker.

Configuration of  FieldsKit ReInterlacer:

  • Field Order: [Lower First]
  • Output Type: [= Create motion estimated fields]
    • This is not the default (oddly).  But it is the only proper way to get the expected “perfectionist” reinterlacing to happen!
  • Source Layer: [Video 1]

Supplier’s website:


Chroma Upsampling (Chroma Interpolation)

Friday, August 31st, 2012

Shooting green-screen onto a 4:2:0 chroma-subsampled format, intending of course to use it for chroma-keying.  Obvious disadvantage is green-ness of green-screen only gets sampled at quarter-resolution.  Not a show-stopper, given my target deliverable is standard definition, but anyhow, towards perfectionism, is there any way to up-sample to 4:4:4 i.e. full definition colour?

It does occur to me that something more sophisticated than chroma blur ought to be possible, broadly along the lines of edge-following methods employed in resizing. What’s out there?

  • Simplest method, that most people seem to use, is chroma-blur.  That’s only the chroma, not the luma.
  • Searching around, Graham Nattress has analysed the problem and seems to have produced a more mathematical approach.  But it’s only available (at time of writing) for Final Cut (which of course is Mac-only at present).

Some tools that “promise” upsampling, but I wonder by what methods:

  • GoPro-CineForm intermediate.  The codec settings include an option to up-sample to 4:4:4
  • Adobe Premiere, but only if a Color Corrector effect employed.
    • But the crucial thing here, regarding the usefulness of this, is whether it uses any better method than chroma blur.

Some questions:

  •  Does Adobe have anything built-in to do something Nattress-like nowadays?
  • DaVinci Resolve?
  • Boris?


Canon C300 – Great Reviews & Posts

Thursday, December 1st, 2011

The Canon C300 is of interest to me as a potential “workhorse” video camera (replacing my current EX3) for both cinematic projects and live events (incidentally, after having written that sentence, I reassuringly found the same phrase uttered by Philip Bloom, so hopefully I’m on the right track here).

Cinematic projects in particular can benefit from more controllable DOF and both can benefit from light sensitivity, while the live events (indoor or outdoor) in particular can benefit from latitude.  Of course these things are handy in general, but those are the occasions when I’ve felt they were lacking in the past.

The C300 addresses most things, apart from lack of 10-bit output.  Hopefully I could trade-up to that in a couple-or-three years’ time when Canon upgrade to that.  And investing in “glass” (lenses) for it is probably a better investment than external recorders (buy or hire) for my current EX3.  The price stings a bit though, so I won’t just leap into it.  I’ll definitely begin by hiring/renting.

My Canon C300 research-in-earnest begins here with a (great) review I initially heard about (via private IOV forum).  That’s the UK’s Institute of videography by the way.  Here:

The following thread at DvInfo includes rolling-shutter-provoking tests (flash, jerk-motion):

The main points about the camera (for me at least, and not in this particular order) are:

  • Ergonomics / practicalities:
    • Better (some say) than the F3.
    • Weatherproof, sensor-cooling
    • ND Filters (three) in-camera
      • Seems fairly unique in this kind of camera
    • Handy proper buttons for Zebra, Peaking, Magnified views.
    • LCD and button-panel orientate in various directions.
    • Fits on a standard DSLR rig.
      • But Zacuto supply a tailored rig.  CVP are among its agents.
      • And there’s Redrock’s inspiringly-named UltraCage.
    • Solid construction, feel and mounting threads, better than “single central bolt” like many cameras have.
    • Wi-fi adaptor transmits a low frame rate version to your computer. If the lens is set to autofocus, you can actually change the focus remotely.
      • Wifi controller is an additional item – not part of the basic package
    • Battery lasts 5 hours
  • Sensor
    • Latitude (recordable, depending on settings):
      • Quoted as “13+ stops in the field”
      • But there is uncertainty over this, since apparently <<Canon thinks that with their Canon-Log color space, the camera allows “800% overexposure… which translates to …an Exposure Latitude of 12 f-stops.” Graeme Nattress of Red disagrees.>>>
    • Less noise, moire and jello than 5D Mk.2 etc.
      • Noise & moire reduction largely result from 4K sensor + DSP to HD.
        • Moire tends to result from significant interpolation inherent in less dense sensors
    • The “less noise” hence less need for low fstops (coupled with smaller sensor than 5D) means greater DOF (the antithesis of the 5D).
      • Greater DOF is preferable at times, for practical as well as aesthetic reasons (when you want to see the background).
  • Lens Mount
    • Camera comes in two lens-mount varieties, not interchangeable:
      • EF-Mount for Canon lenses (ordinaire & “CN-E”).  This is the one for mortals like me.
        • Permits iris to be controlled via dials on camera.
        • Greater available selection of lenses (also usable on stills cameras) e.g.:
          • Telephoto: “EF 70-200mm f/2.8L USM” telephoto zoom lens and tilt-shift lenses.
          • Macro: “EF 100mm f/2.8L Macro IS USM” (article photos include an image of ants)
          • Fisheye: “EF 8-15mm f/4L Fisheye USM”
          • TiltShift: “TS-E 90mm f/2.8”
        • Image Stabilisation (IS) is reported to be good e.g. with the Canon L 70-200mm F2.8 IS Version II
      • PL-Mount: More for movie professionals.
  • Recording:
    • Resolution: HD 1080p & 720p
    • Frame-rates: variable 1 to 30 fps in 1080p mode, and 1 to 60 fps in 720p mode, 1 fps increments.  Also time-lapse and stop-motion/claymation (latter is several frames per “click”).
      • But, like F3 (and unlike FS100), overcranking requires dropping to 720p
    • It can also shoot 50i and 60i (interlaced), useful for deriving 50p and 60p in post.  Historically (e.g. for DV or HDV (Z1) footage), I have done this using freeware (AviSynth and its TDeint filter) for this, but Apple Compressor and other alternatives exist too.  Stu Maschwitz apparently covers this topic in his book << The DV Rebel’s Guide>>
    • Gamma: includes Canon’s Log Gamma.  LCD & V/F display flat and corrected.
    • Audio: uncompressed 16-bit audio at 48 khz (info from FAQ).
  • Recording format:
    • Compact Flash (over 5 hours for a 128GB card)
    • MPEG-2 Long GOP 4:2:2 MXF codec with a constant bit rate of 50 megabits/sec.
      • Philip Bloom says “It is the bare minimum for HD acquisition, but it at least reaches it.”.
      • Sample recording (MXF as stated) is available here (according to here).
      • {Does MPEG-2 imply 8-bit?}
        • Yes.  That is highlighted in several other sites e.g. this and this.  Some believe (rightly or wrongly) that this limitation is a “corporate rather than technical” decision, and a future generation will have 10-bit.
      • {What of the Log Gamma? 8-bit (is it ?) would limit its usefulness or not?}
        • Opinions differ.  Obviously the extent to which it matters depends on the scene.  Some views and image comparisons are here, for 10, 8 and even 7 bits.
          • At that link, one poster suggests dithering as a work-around to reduce banding (from any camera): adjust camera to give noise then (in post) use a good noise-reduction plugin.
      • The only Super-35 sensor camera in its price category (as of fall, 2011) that records 4:2:2 color sampling on-board.
  • Ports:
    • XLR (via clip-on monitor assembly?)
    • Time code, Genlock, HD/SD-SDI and HDMI
      • {But how many bits? 10 (as I’d hope) or still only 8?}
        • Only 8-bits, surprisingly.  It is reported that: “SDI output (is) limited to 8-bit 4:2:2”
  • Monitoring
    • RGB histogram, vectorscope and an Edge monitor (focussing)

More links:

Other Misc. Links:

Moire pattern (shirt) fix in Sony Vegas

Wednesday, December 8th, 2010

Moire pattern on presenter’s shirt when footage’s image size shrunk from HD to much smaller.  Was a corporate shoot where lecturer wore a finely lined shirt (doh!).  Tried a few experiments but a googled solution worked best, as follows.  I applied a Gaussian blur as a MediaFX, so it (hopefully) got applied before the pan/crop downscaling.  Worked well in any case.

    • Your problem is moire. If you want to understand why you have this problem Google “Optical Low Pass Filter”. This is a vital part of any digital camera both still and video. When you downscale an image to a lower resolution the OLPF is no longer enough to filter out high frequency detail at the lower resolution and hence you hit problems with the Nyquist Limit.
    • You fix this by reducing the resolution BEFORE downscaling. This is easily done by using the Gaussian Blur FX in Vegas. You only usually need a tiny amount and can adjust the amount depending on what is in front of the camera. Values of 0.001 to 0.003 should be more than adequate. You may only need to apply it in the vertical direction. …you will need to experiment to get just enough image softening to stop the aliasing without making it too soft.
    • …you need to do this BEFORE downscaling. Render at Best quality mode. Do some quick tests

DCT (Mpeg/Jpeg) Gibbs Noise / Mosquito Noise

Monday, February 8th, 2010

The Gibb Effect is an MPEG compression artefact. It is a blurring of the outline of sharp objects, with inappropriately-coloured pixels appearing around the outline of the object. The commonest area in which the Gibb Effect is seen is during end credits, where insufficient bits have been allocated to compressing the data. Another name for this artefact is mosquito noise.[]