[opendtv] Re: The curse of Bayer pattern sensors

  • From: Alan Roberts <roberts.mugswell@xxxxxxxxxxxxxx>
  • To: opendtv@xxxxxxxxxxxxx
  • Date: Sun, 09 Feb 2014 13:24:29 +0000

There have been many developments in this field over the last 20 years, and international conferences concentrating on just this subject. Every manufacturer seems to have their own solution, some are reasonably good, but most are rather poor. I test cameras for a living, and it's rather amazing what you find when you start investigating products properly.

Canon's solution in the C300 is the best use of the Bayer pattern (4 photos-sites/pixel, one R, one B, 2 G) but it's overkill. ARRI have just about optimised it in the Alexa (using 2880x1620 to make 19820x1080 with 1.5:1 down-scaling containing the Bayer decoding).

Some manufacturers are experimenting with different patterns, usually to raise the sensitivity at the expense of coloured aliasing. The most straightforward solution is the RGB vertical stripes as a triplet per pixel, Sony F65, buty even that has problems.

By the way, my long absence from here is simply because the vast majority of what's discussed here is of no interest to me at all. Just occasionally something interesting arises, and then ...

Alan Roberts

On 09/02/2014 12:40, Olivier Houot wrote:
Sorry for bringing up such a prehistorical concept, then, but i was
surprised by the fact, and thought other people on the list might be
unaware of it, too.

But then do you know how manufacturers usually work around the problem ?
Do they optimize for green, red/blue, or something in between ? (or
perhaps no low pass filter at all except lens natural MTF, and so much
the worse for sampling theory).

P.S. nice to see so many people are still monitoring Opendtv, i was

You can UNSUBSCRIBE from the OpenDTV list in two ways:

- Using the UNSUBSCRIBE command in your user configuration settings at FreeLists.org
- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word 
unsubscribe in the subject line.

Other related posts: