Thursday, 11 December 2008

Re-colouring the past

A bit of a non-bibliophile and vanity post (I'll get to that bit) but it does spring from something currently in print. This week's Radio Times - 13th-19th December 2008 - has an interesting article by Jon Wilde, "Re-colouring the past", about the recolouring of "Room at the Bottom", a 1969 episode of Dad's Army. The article's not online, but there's an equivalent at The Guardian (Unscrambling an army of colours, Charles Norton, 11 December 2008). The restored episode is to screen on Saturday 13 December, BBC2, 8.25pm.

The RT's explanation of how this was achieved wasn't quite clear, but I was interested to find source, the Wiki of the BBC's Colour Recovery Working Group. The section An examination of source material explains that the episode was originally filmed in colour video format, but archived to monochrome 16mm film by "telerecording" (i.e physically refilming by pointing a movie camera at the colour output on a big TV screen) and only the archive remains. Sometimes, however, telerecording was done without subsequently filtering the colour signal, so the mono film shows the position of the RGB phosphor dots on the colour screen as an overlay of artifacts ("chroma dots") whose pattern contains partial information about the original colour signal. With plausible assumptions, this makes digital colour recovery possible. Techie details here and here.

The actor Ian Lavender stresses in the RT article that this is restoring material to its original coloured condition. He expresses the common view that the more radical process of colourising originally b&w film is "cultural vandalism" (a phase previously used by the Writers Guild of America West on the same subject - see the Museum of Broadcast Communications). Others think it legitimate creation of novel value-added material. In between those extremes, accurate restoration of originally-coloured material seems pretty uncontentious.

This leads to the vanity bit: a few years back I wrote an article which may be of related interest on the whole topic of historical photo and video colourisation: The colour of the past (Scientific Computing World, July/August 2005). It's not just computer-geeky, I promise: there's plenty of historical interest such as Sergei Mikhailovich Prokudin-Gorskii's remarkable early colour photos (see The Empire that was Russia for a taster); Don P Mitchell's reprocessings of the equally remarkable (and largely forgotten) Venus images from Soviet landers; Mark Simpson on "The Colour of Whiteness" (the cultural effects of the development of colour photography); and a box-out by The Growlery's Felix Grant on the world's oldest known photograph, Joseph Nicéphore Niépce's iconic image View from the Window at Le Gras, c. 1826.
- Ray


  1. This is really interesting. However, I don't think it will ever be possible to reproduce the color of the light. All of the attention seems to be on the color of objects/surfaces, when it is the light (blue-sky, yellow-sun, green-woods, or whatever color may be reflected from any/all large nearby surfaces) that "makes" the particular tint that you see. Within any given scene, there will almost always be multiple tints (yellow tint from the primary with blue tint from the secondary [shadows] is common, but there can be any number of secondary sources). Wherever one light source is blocked (shadowed), the object will be lit by the secondaries which are almost never going to be the same tint as the primary.

    With rounded objects (which is just about everything) you have gradient/gradual transitions from one light tint into another (light into shadow in the simplest instance).

    And, of course, all of these light tints are changing every second if they're in a motion picture.


  2. I don't think it will ever be possible to reproduce the color of the light.

    If you mean it's difficult to back-engineer the particular illumination conditions and white-light colours of objects that resulted in the particular appearance of objects, yes. (That is, you see a pink piece of cloth on screen, and there's no way to tell, say, if it's pink cloth under white light, or white cloth under pink light).

    But that seems academic in this case, in that they're just interested in recreating what was stored on the PAL colour video (i.e. what was "seen" by the camera) regardless of whatever multiple sources went into producing that).

  3. Me, I'm just admiring your new "Planet of the Apes style" portrait...


  4. Generated using the University of St Andrews Face Transformer.