Interesting Leica product announcements today ...

Discussion in 'Digital Cameras' started by Bruce, May 10, 2012.

  1. Precisely the circumstance in which I was suggesting it was a useful
    thing to do.
     
    Chris Malcolm, May 19, 2012
    1. Advertisements

  2. This particular kind of noise reduction (subtraction of an
    equivalently exposed black image) seems to me on theoretical grounds
    to be more likely to improve image quality than reduce it. It's
    completely different from the kind of noise reduction processing used
    in translating RAW images to jpegs in camera or RAW image processor,
    which effects a compromise between noise reduction and detail loss.

    Looking at the results in practice, same tripod photograph with it
    switched on and off, it seems to me quite definitely to improve IQ,
    even to a determinedly pixel-peeping scrutiny.
     
    Chris Malcolm, May 19, 2012
    1. Advertisements

  3. Bruce

    Mxsmanic Guest

    As I recall, it is derived from microfilm. Intended to produce extremely high
    resolution images with no grayscale.
     
    Mxsmanic, May 19, 2012
  4. Bruce

    Mxsmanic Guest

    That's because it actually uses informmation from two separate images to form
    a single image, so it's possible for it to reduce noise without degrading the
    target image significantly.

    However, noise reduction performed on a single image--without a reference
    image to provide extra information--will always degrade image quality. It's
    mathematically inevitable.
     
    Mxsmanic, May 19, 2012
  5. Bruce

    John A. Guest

    The idea is that any non-random noise sources will negate each other
    in the two exposures, while random noise will tend to average out as
    well.

    A similar approach would be to take two or more exposures and average
    the pixels together, provided you can get the registration right. In
    theory, the average of a number of exposures approaching infinity
    should be spot-on with zero random noise. How many it would take to
    get the expected noise below the color-depth-imposed threshold would
    vary with how much noise you're dealing with I would imagine. Though
    of course there can always be outliers.

    I recall many years ago my dad had a box he could plug his video
    camera into that did that with video frames to produce a still that
    was of much better quality than and of the constituent frames.
     
    John A., May 19, 2012
  6. Not necessarily. You can often find reference patches within the same
    image - areas which you know to be of a near constant brightness for
    example - and measure the statistics of the image there. The fact that
    noise reduction (for viewing by humans) can work so well shows that
    perceived image quality can be improved rather than degraded.

    David
     
    David J Taylor, May 20, 2012
  7. Bruce

    Noons Guest

    David Dyer-Bennet wrote,on my timestamp of 20/05/2012 5:14 AM:
    Like all the other Canon fan-boys. And any maker fan-boys.
    Nothing but groupies. Rarely any substance or recults to show what they are
    capable of. But oh-boy: do they know how to dump on others...
     
    Noons, May 20, 2012
  8. Bruce

    Noons Guest

    David Dyer-Bennet wrote,on my timestamp of 20/05/2012 5:12 AM:
    I said "similar Pan films"...
    The image characteristics and formation for the techpan look alikes is
    completely different from the traditional b&W stuff. Negative strips examined
    on the microscope are completely different: there is no "clumping" anywhere.
    Its original design purpose was as a micro-film.
    The biggest difference I see between all these high-res films and the run of the
    mill tri-x/pan-x/t-max is that the non-exposed portion of the negative comes out
    totally transparent instead of with the usual grey "fog". To the point where
    Adox CMS20 for example can be used as a b&w "slide" film with contact
    duplication: plenty of folks at APUG have tried it and love the results. Same
    for the Rolleis. It also gives the negatives a huge dynamic range that exceeds
    what my ED9000 can do almost every time. That makes it a challenge to scan a
    high dynamic range image in its entirety without losing something off the
    "ends". HDR-like scanning works to "compress" it, but it's a RPITA time-waster...


    Who doesn't? :)
    I'm very much into bracketing. With the F6/F100 and spot meter I tend to stick
    to the zone system. But with the dslrs and the Oly m4/3, bracketing is the
    order of the day. Although I must admit I'm getting it right most of the time
    first go with the Oly. Then again, I use it a lot more than the dslrs...
     
    Noons, May 20, 2012
  9. Bruce

    Mxsmanic Guest

    If you can trust that portion of the sensor to be representative of the rest.

    Often sensors have irregular noise across the sensor. And sensors can vary
    greatly from one camera to the next. I've seen two cameras with consecutive
    serial numbers that had huge differences between them: one had no visible
    noise at all, the other was just a mass of noise.
     
    Mxsmanic, May 20, 2012
  10. Perfect description of Noons.

    -Wolfgang
     
    Wolfgang Weisselberg, May 20, 2012
  11. [/QUOTE]
    Random noise (if observed statistically) will grow with the square
    root of measurements, as it's a random walk.
    Actually, the total noise grows (see above), but much slower than
    the total signal. Thus, the SNR goes to infinity.

    As the main noise source is photon noise, you could just expose
    longer (and thus add the inevitable read noise less often).
    Of course, after some point your pixels will overflow, but who
    really cares for such real-life limitations, when one can just
    postulate infinitely deep electron wells?

    -Wolfgang
     
    Wolfgang Weisselberg, May 21, 2012
  12. Candles & fire in general, incandescent light, "warm white"
    lamps, --- nearly all your typical indoor lighting falls under
    that heading.
    Even if you are going to make an imbalance in post-processing
    --- starting from a balanced point increases your reach and/or
    your quality.

    -Wolfgang
     
    Wolfgang Weisselberg, May 21, 2012
  13. So let's see how the tools reconstruct that:
    http://upload.wikimedia.org/wikipedia/commons/f/fb/Moire_pattern_of_bricks_small.jpg

    While we're at it, the luminance pattern shouldn't be there
    in that textile:
    http://ueberlicht.de/wordpress/wp-content/uploads/2011/10/Luminanz-Moire-Stoermuster-Stoff.jpg

    -Wolfgang
     
    Wolfgang Weisselberg, May 21, 2012
  14. Bruce

    Trevor Guest

    I've covered that here in the past, YOU may want your candle lit photo's
    etc. to look like they were shot in daylight, but I certainly don't, so no
    filtration is necessary for me.

    You mean *IF* you are going to make an imbalance in post, *not* "even if".
    Deliberately making the color balance wrong, when it is actually what you
    want to start with, is pointless.

    Trevor.
     
    Trevor, May 21, 2012
  15. That sounds like a faulty camera - or different default settings.

    In practice, it works well in the version of Paint Shop Pro which I have.

    Cheers,
    David
     
    David J Taylor, May 21, 2012
  16. Bruce

    Alan Browne Guest

    Candle lit is one thing.
    But one rarely wants incandescent coloring and less so fluorescent coloring.
     
    Alan Browne, May 21, 2012
  17. Bruce

    Trevor Guest

    Actually I do usually want incandescent coloring or I don't use them. And
    try to avoid fluoro lighting too, and usually do. But funny that fluorescent
    wasn't one you mentioned anyway! However I do have a fluoro filter in my
    kit, but since fluoro's vary so much, I find it easier to adjust in post
    these days anyway. Try finding a fluoro filter that exactly matches a quad
    phospur tube for example, good luck! :)

    Trevor.
     
    Trevor, May 22, 2012
  18. You're completely missing the point. Please read more carefully,
    I am not, repeat, NOT saying your candle lit photos should look
    like daylight.

    However, with candle light your blue channel is *severely*
    underexposed when your red channel is already at capacity.
    To fix that, you need to reduce the light hitting the red pixels
    very much and the light hitting the green pixels somewhat.[1]
    I.e. you need a filter.

    Of course you are free to use the white balance tools in post to
    re-create the very red environment.

    There is no "wrongW balance. Digital isn't film.
    Proving that you didn't grasp anything at all in this thread,
    but wanted to jump on someone to prove you're better.

    -Wolfgang

    [1] Or you need to add green and blue light, e.g. with a
    flash.
     
    Wolfgang Weisselberg, May 22, 2012
  19. Bruce

    Alan Browne Guest

    Not sure why you would say that as I haven't mentioned lighting in this
    thread at all.
    There is no fluorescent filter (glass or in PS) that can completely
    correct for a fluoro, merely shift it off green reasonably.

    One can't avoid shooting in such conditions if one shoots industrial and
    commercial locations. One can correct for them as desired, and except
    for some effects, that light is hideous. Moreso when the building
    managers don't stick to a type (warm v "day", etc.)

    Incandescent, uncorrected, makes for horrid skin tones.
     
    Alan Browne, May 22, 2012
  20. Bruce

    Trevor Guest

    ?????
    "Candles & fire in general, incandescent light, "warm white" lamps, ---
    nearly all your typical indoor lighting falls under that heading"

    seems to mention lighting to me, and ignores flash of course.

    Right, so how does that support your argument about using filters on camera
    being better than post again?

    Right, when flash is not an option.

    Exactly why the best you can do is in post.
    Sometimes, but only when you must use it, and can usually be made acceptable
    at least in post. And sometimes that is what you are after, my favourite is
    to add a little fill flash when necessary, but still look like incandecent
    lights if that's what I'm after.

    Fortunately I no longer use many of my huge collection of filters. People
    starting these days are lucky they no longer need to buy many IMO.

    Trevor.
     
    Trevor, May 23, 2012
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.