Slashdot Mirror
Reverse Engineering Doctor Who Into Color
Lanxon writes "In 1967, the BBC set about junking its Doctor Who archive: a moment sci-fi fans wish they could travel back in time to prevent. There are 108 vintage episodes missing, but since 1978 a number have been rediscovered as 16mm black-and-white films. The BBC shot many of these series in color, but made monochrome copies for countries such as Australia, where many TV companies were still broadcasting in greyscale. The reels had sat in archives since. Now, the Doctor Who Restoration Team, an independent group contracted by the BBC, is using a new technique to regenerate The Doctor in color."
Since this is the BBC, they shot *none* of them in color but many of them in *colour*....
So the article was devoid of anything of particular interest other than some jargon. The jargon, on the other hand, led to fascinating little technique about reconstructing the color of the grayscale image from "chroma dots". The actual method was discovered by a BBC engineer, and you can read more about it here: colour-recovery.wikispaces.com.
Actually, the process by which they're recovering the colour data is very interesting:
http://www.insell.co.uk/colourisation/Recovery_of_Colour_Information_0-2.htm
When you have nothing left to burn you must set yourself on fire
It almost is reverse engineering. The chroma subcarrier in a video signal has a center frequency picked to allow the sidebands to fall between those of the main lumanance (black and white) video. The spectrum of those extends out from the main visual carrier frequency (or up from D.C. for the baseband signal) at multiple of the horizontal scan rate. The goal was to add color broadcast information to an existing greyscale system while introducing a minimal amount of interference. Here people are figuring out what is going on from the visual interference.
The added signal amplitude represents the amount of color added/subtracted from the greyscale white, and the phase represents the hue. The phase of the signal is compared with a short burst (a minimum of eight cycles) sent just after the horizontal sync pulse prior to the start of video on each scan line. PAL, as used by the BBC, is very similar to NTSC, except the scan rates differ, the phase of the reference signal is inverted on every other line to help cancel out the effect of small phase errors on tint.
Basically, those trying to recover color from the back and white films of on-air video have to use a comb filter to pick off the frequency (precisely related to the inverse of the spacing) of the resulting dots that are there from the color signal. The position of the dots from left to right carries the phase information. Considering that the dot pattern is probably quite weak, the resulting color would be noisy. Depending on the filtering used, the bandwidth (detail) may also suffer. But it is still a good starting point to know what the colors were.
The dots aren't on/off like pixels. It's actually a sinusoidal intensity variation. I recall some older Zenith B&W sets had particularly good detail (and maybe some video peaking - enhancement) making it easy to see which programs were broadcast in color, and what parts of the picture were deeply saturated. In addition to a notch in the video response at 4.5 MHz to filter out patterns from the sound, some sets rolled-off or notched centered at 3.58 MHz (3.579545 actually) video response to reduce the interference. Better later sets (and color generally) used "comb" filters to separate the interleaved spectral components without those loss of detail seen with more primitive methods. Failure to filter color signals could cause wild colors/patterns on things like striped neck-ties when a shot zoomed in/out.
It's pleasing to see that there are still a few around that understand the old analog technology well enough to realize there were visual color cues remaining. Even those that understand the electronics well often don't associate a particular visual characteristic with the responsible signal attributes.
Although partial signal recovery is easy to envision with analog electronics, something along the lines of a GIMP/Photoshop plugin could work as well. Some might think of it as being similar to watermark detection.
Many people like Pyramids of Mars, and the Talons of Weng Chiang, though the latter isn't particularly culturally sensitive.
I agree, but there's a wonderful moment when Tom Baker exclaims something like, "Wait a minute, you're Chinese," as if that visually obvious fact had eluded him up to that point. Made quite an impression on my young mind, that an alien -- even a super intelligent one -- would be less capable of seeing our trivial differences. To be truly unprejudiced, we must see through better eyes.