Musings on Photography

Processing power for a new metering system

Posted in equipment by Paul Butzi on May 4, 2007


In this thread on the Canon 10-d discussion group (registration required, sorry), J Bryan Kramer writes 

Thinking on this I suspect that he is under estimating the processing power required to duplicate his mental processing about exposures. The requirements may be more than is currently practical for a camera cpu. But still, it seems that they could improve the in camera metering.

I don’t think I’m underestimating the amount of processing.  Finding the brightest pixel (and the darkest pixel) isn’t a task on a par with generating a jpg from the raw data.  Just keep the lowest value and the highest value as you read the data out of the sensor.  Remember, you don’t care where the brightest/darkest pixel is, you just care how bright/dark it is.

A bigger problem would be positioning the sensor in the optical path so that it can see the full frame without impacting viewfinder brightness substantially. 

But the biggest problem is, I suspect, that the engineers are trapped in an old paradigm.  Things have changed, and camera design needs to change to keep up.  The best possible digital SLR is not just a really great film SLR with the film transport ripped out and a digital sensor duct taped into its place.

5 Responses

Subscribe to comments with RSS.

  1. chuck kimmerle said, on May 4, 2007 at 7:27 am

    I think that the camera meter of your dreams would have to be far more intelligent than just finding the brightest pixel and giving it a value of 254 (as an example). For instance, since most specular highlights are anomalies to the ambient scene lighting, a meter would need to be able to distinguish, and thus ignore, them in exposure calculations or risk severe underexposure.

    As well, since there may be more than one correct exposure for any one scene, depending on the photographer, there will always be someone who is not happy.

    I think, though, that you are right in your position that meters can be made more accurate, but I think it’s far more difficult and involved that you may anticipate.

  2. Guy Tal said, on May 4, 2007 at 7:45 am

    “A bigger problem would be positioning the sensor in the optical path so that it can see the full frame without impacting viewfinder brightness substantially.”

    Take a peek at the new 1D Mk III – an SLR with live view on the LCD. Seems like much greater feats are already a reality.


  3. Martin Doonan said, on May 4, 2007 at 11:13 am

    Chuck – with the spatial resolution Paul suggests earlier, I don’t think speculat highlights would throw the sensor off. Metering pixels larger than sensor pixels.

    Guy – if I’ve read correctly, the 1DIII needs mirror up for live preview (which it then proceeds to lower before taking the shot!).

  4. matt said, on May 4, 2007 at 11:31 am

    Did I miss why you need another sensor? There’s a matrix meter already built into the camera, but it works on a corpus of scenes that assume we care about exposing for midtones and shadows. Seems like all you would have to do is replace its corpus of scenes with one that assumes we want to protect highlights at all costs.

  5. J bryan kramer said, on May 4, 2007 at 4:05 pm

    LOL, I’ve gone to shutter clicker heaven, I’ve been quoted by Paul…heh. Or chastized by Paul I suppose.

    Actually your succeding post answered all my questions and I have to say I’m in full agreement with you now. I was meaning to post a link to that subsequent post: ‘What would a decent metering system look like’ on Canon-10d but have been lazy and have not done it yet.

    I’ve been posting links to your Art posts to Artshow-photo and stirred up a bit of talk over there. You provide all sorts of linking material Paul, all thought provoking.

Comments are closed.

%d bloggers like this: