Dummy question about CCDs and noise

Discussion in 'Digital Photography' started by J.S.Pitanga, Nov 5, 2004.

  1. J.S.Pitanga

    J.S.Pitanga Guest

    Hi there,

    It is known that a higher amount of pixels per area of CCD tends to
    produce a higher amount of noise. Thus, a 3mp 1/2.5" CCD would produce
    less noise than a 4mp 1/2.5" CCD.

    My question is whether a camera with a 4mp 1/2.5" CCD but set to a
    resolution 3mp produces an amount of noise typical of its 4mp CCD or a
    lower amount of noise, such as that of a 3mp 1/2.5" CCD?

    Thanks in advance for any answers,

    J.S.Pitanga
     
    J.S.Pitanga, Nov 5, 2004
    #1
    1. Advertisements

  2. The size of the light collectors is one factor but there are many other
    factors that dominate. Sensor technology still has some flaws.
    Downsampling will eliminate the highest frequency noise by averaging it
    out, the same as if you blurred the image. It doesn't fix the patchy
    CCD flaw patterns that were common on 4MP sensors a few years ago. It
    also doesn't fix the blotchiness caused by the noise filters required on
    inexpensive sensors.

    There are better methods for removing noise. Some utilities will
    analyze the frequencies prominent in the noise and then subdue them. It
    causes the blotchiness when the subject of your photo is mistaken for
    sensor noise (trees and lawns are common victims) but it's better than
    simple blurring.
     
    Kevin McMurtrie, Nov 5, 2004
    #2
    1. Advertisements

  3. J.S.Pitanga

    Bob Williams Guest

    The noise will be the same at 4 or 3 MP.
    Your camera CAPTURES all 4 MP, regardless of your setting.
    After capturing, the camera will resample the image downward to 3MP and
    then SAVE the resampled image on your memory card.
    Bob Williams
     
    Bob Williams, Nov 5, 2004
    #3
  4. I disagree with the other people posting here. Averaging pixels (by
    resampling down to a smaller image) *does* reduce noise by averaging out the
    different fluctuations of different pixels.

    In astronomical imaging it is common to "bin the pixels 2x2." That means
    turn 4 pixels into 1 by averaging. Your 4-megapixel sensor then delivers a
    1-megapixel image, but it's a much less noisy image.
     
    Michael A. Covington, Nov 5, 2004
    #4
  5. J.S.Pitanga

    Mitch Alsup Guest


    In the end it comes down to the Full well capacity of the sensor cells
    and the readout noise of the amplifier and A/D converter (with a minor
    contribution of the 1/F noise of the readeout frequency).

    If the 4MP sensor has the same full well capacity as the 3MP sensor,
    and the readout noise is comparable, then there will be no essential
    difference in noise. However, full well capacity tends to be smaller
    in bigger MP sensors of the same size as smaller sensors.

    As an example: take a look at the 6MP Canon 10D versys the 8MP Canon
    20D. These have the same sensor size, but better on sensor electronics
    have conpensated for the smaller sensor cell and at low ISO levels
    both cameras have essentially the same noise levels. At higher ISOs
    the better on-die and readout electronics and lower operating speed
    (4 channels at 16 MHz versus 2 channels at 24 MHz) of the readout
    electronics enable the 20D to have better noise than the 10D.
     
    Mitch Alsup, Nov 5, 2004
    #5
  6. One must not make a blanket statement on this. It is one of those "all
    else being equal..." things. The area density (actually, the area of
    each pixel) is only one factor in the total noise of a detector. And,
    we must be careful to differentiate original detector noise from the
    final 'noise' in the output file. Post processing CAN reduce noise,
    but it depends on the algorithms used. I suspect the difference in
    the two cases below would not show a dramatic difference in noise
    either way.

    The final noise in an image is a result of inherent detector noise,
    preamp noise, quantization noise, processing algorithm noise, and the
    phase of the moon.
     
    Don Stauffer in Minneapolis, Nov 5, 2004
    #6
  7. J.S.Pitanga

    Jim Townsend Guest

    It's the *size* of the pixels (or image sensors on the chip that matters).

    Consider placing two bottles out in the rain. In one of the bottles
    you insert a large funnel.

    After a period of time, the one with the large funnel will always have more
    water in it due to the larger capture area.

    Same goes for camera sensors. Their ultimate job is to produce a voltage
    that's proportional to amount of light falling on them. A large sensor
    can gather more light and produce a higher voltage.

    The higher voltage produced by a large sensor requires less amplification to
    be useable. Since it's the amplification that produces most of the noise,
    less amplification means less noise.

    If you change the resolution in your camera, you're still using the
    same sized sensors, so the resulting noise will remain the same.
     
    Jim Townsend, Nov 5, 2004
    #7
  8. J.S.Pitanga

    usenet Guest

    In general, yes, that's correct, & a useful rule of thumb. But you need
    to remember that it relates only to the theoretical limits of a silicon
    photosensor only. There are many other factors involved besides the
    photosite size that also affect noise levels, such as ambient heat, the
    quality (ie $$$, skill & creativity) that went into the design of the
    sense-amps, conditioning circuitry, A2D converter & other support
    electronics. Then there's the digital processing by the cameras onboard
    computer, which typically includes software noise-reduction & other
    enhancements.
    (These sorts of factors are why a Brand 'X' 4MP, 1/2.5" sensor camera
    might have very different noise levels to a Brand 'Y' camera that has a
    sensor of the same size & resolution, & why, (for a recent example),
    Canon were able to make the EOS 20D, 8MP DSLR *less* noisy than the 10D,
    despite having about 25% more pixels on the same size sensor.)
    Your question is not even slightly dumb. It's actually very perceptive.
    :)

    In theory, summing data from multiple photosites should indeed reduce
    noise somewhat, much as it would if the sensor had the smaller number of
    photosites, with the equivalently larger area. And most of the time, it
    works in practice, & you'll find (unless your camera is poorly
    designed), that high-ISO images will look less noisy if taken at smaller
    resolutions.

    So, the obvious interpretation of this relationship would seem to imply
    that if you're finding noise to be a problem in your photos (eg;
    available light shooting at ISO 400+ on a digicam), & want cleaner
    photos, it'd be helpful to shoot at a lower resolution than that of the
    sensor itself.
    *However* - you don't need to do that, because the maths that governs
    the noise vs photosite-size tradeoff on the sensor also applies to /any
    stage/ between the light hitting the photosite & you printing the
    resulting image on your inkjet or at your local lab. This means that,
    all else being equal, you can also reduce noise in your images by simply
    reducing the resolution of your image after you've taken it. This means
    that you lose nothing (except disk space), by shooting at the native
    resolution of your camera, & simply reducing the size of the image in
    Photoshop (or any other program that scales with a high-quality
    algorithm).
    You can easily demonstrate this for yourself by taking a photo at your
    cameras highest resolution & ISO setting, then again at a much smaller
    size (eg; 50%), then resampling the large image in Photoshop down to the
    exact same resolution. You will most likely find that, (again, all else
    being equal, etc), both images will have similar amounts of noise.

    So, to rephrase your original question: Is a 3MP image from a 4MP
    sensor going to be less noisy than an otherwise identical 4MP image from
    the same sensor? - The answer is most likely 'yes', but not by a whole
    lot, due to the final image being nearly as big as the sensor. Also, it
    is unlikely to give you any advantage (in terms of noise levels), over
    just shooting at 4MP & resizing to 3MP on your computer.
    My pleasure. I hope you found my comments informative. :)
     
    usenet, Nov 5, 2004
    #8
  9. Downsizing reduces noise, so it should be closer to tha from a 3mp CCD
    than a 4mp CCD, all things being equal.

    -Tripurari
     
    Tripurari Singh, Nov 6, 2004
    #9
  10. The camera does capture at 4mp, but downsizing reduces noise by
    averaging pixel values.

    -Tripurari
     
    Tripurari Singh, Nov 6, 2004
    #10
  11. J.S.Pitanga

    Bob Williams Guest

    But it also degrades the image.
    There is NO WAY to increase the signal-to-noise ratio of an image.
    If there was, you could continue using it until the image was sharp as a
    tack and noise free.
    Like a free lunch......it never works out that way.
    Bob Williams
     
    Bob Williams, Nov 6, 2004
    #11
  12. There are many PS filters, actions, plugins and standalone apps that
    decrease noise, thus increasing at least the apparent ratio.
    [If there were, but there is!]: It doesn't follow that you get both
    noise reduction and sharpness. Nor does any beneficial adjustment work
    on a straight line or even a favorable curve. At some point the
    tradeoffs become quite unacceptable.
    Amen to that!
     
    John McWilliams, Nov 6, 2004
    #12
  13. Downsizing trades off resolution for noise, I see no contradictions
    here. Nor do I see a way to get clean sharp image by repeated
    application.

    Coming back to the OP's question, if you downsize a 4mp image to 3mp
    you'll get the resolution and noise of 3mp sensor.

    -Tripurari
     
    Tripurari Singh, Nov 7, 2004
    #13
  14. J.S.Pitanga

    J.S.Pitanga Guest

    Dear listmembers,

    Thank you so much to all of you who so thoroughly and kindly answered my
    question, thus shedding a lot of light on my noisy brain sensors!

    J.S.Pitanga
     
    J.S.Pitanga, Nov 7, 2004
    #14
  15. J.S.Pitanga

    Big Bill Guest

    Isn't it also a much less detailed image?

    Bill Funk
    Change "g" to "a"
     
    Big Bill, Nov 7, 2004
    #15
  16. Of course, assuming the image, optically, had enough detail for it to
    matter.
     
    Michael A. Covington, Nov 8, 2004
    #16
    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.