fake resolution, and JPG compression in digital cameras -- how to measure?

Discussion in 'Digital Cameras' started by Ted Shoemaker, Jun 8, 2011.

  1. Hello,

    I've been reading about effective resolution, interpolated resolution,
    and related stuff.

    We recently bought a low-end model of a good-name camera (Canon
    A490). It takes 10-megapixel pictures, but then saves them into JPG
    files that are (so far) between 1.3 and 4.1 MB. That's a lot of
    compression.

    In fact, this means that my 10-megapixel picture is no sharper than a
    much smaller picture could be. I'm crudely guessing that the "true
    resolution" is closer to 2 megapixels. (I would use the term "image
    quality" but that sounds like it can refer to other, and more
    subjective characteristics.)

    A few questions:

    1. How can I measure the "true resolution" that my camera produces?

    2. Is there a website or consumer guide that publishes the "true
    resolution" of cameras? Next time, I'd rather know before buying.

    3. Can I tweak the settings on my Canon A490 to take pictures with
    higher "true resolution"?

    Thank you!

    Ted Shoemaker
     
    Ted Shoemaker, Jun 8, 2011
    #1
    1. Advertisements

  2. Ted Shoemaker

    ray Guest

    I believe you know the 'true resolution'. By default the camera processes
    to make a jpg image. In so doing, detail rather than resolution, is lost.
    Does chkdk do raw images? According to what you say, it sounds to me like
    that is what you are looking for. Do you have 'quality' set to 'high' or
    'best' or whatever is offered? If so, that's as good as you can do
    without a firmware change (chkdk for canon cameras). The other option
    would be to get a camera that supports saving RAW files. The ability to
    save in RAW format was one of my criteria when I selected my Kodak P850
    several years ago. Some point and shoot cameras have that ability - AFAIK
    all digital SLR cameras do.
     
    ray, Jun 8, 2011
    #2
    1. Advertisements

  3. Ted Shoemaker

    ray Guest

    BTW jpg is a lossy compression method - meaning that detail is lost when
    it happens. Not all compression methods are like that. There are quite a
    few lossless methods as well, though none of them work all that well on
    most photographs - due to the entropy inherent in digital photographs
    (the same is true, for example, of compressing computer executables).
     
    ray, Jun 8, 2011
    #3
  4. By photographing a resolution chart. Then you look at the picture,
    and find
    the finest group of lines that you can distinguish, and read the
    number
    by them, and that's the resolution (in that part of the frame; the
    lens
    doesn't produce the same resolution all across the image, though
    the rest of the system is usually "flat"). (Chart must be used
    according to instructions to produce valid results.)
    No, it's not a concept that's in general use.
    I think you're making a bunch of incorrect assumptions, and also that
    you're privileging resolution over other image characteristics that
    are at least equally important.

    But let's step back. WHY do you want to improve the resolution
    of your photos? Is this just a theoretical desire, or are there
    specific aspects of how your photos look that you think need
    fixing, and have decided resolution is the thing you need to
    increase?

    (Or is this just general scientific curiosity? Always a good thing,
    but in photography I tend more towards practical image-making,
    and can't reliably address the fine scientific details of a lot
    of things.)
     
    David Dyer-Bennet, Jun 8, 2011
    #4
  5. Ted Shoemaker

    isw Guest

    It's probably a lot more accurate to talk about JPG "encoding" rather
    than "compression"; it is a form of "perceptual coding". True, the files
    are smaller, but not as a result of just blindly reducing resolution (or
    anything else). The algorithm is quite sophisticated, and it's intention
    is to replace parts of the image with things that look the same to your
    eye but take up a lot less storage. Done properly, it can be pretty
    difficult to tell.

    Isaac
     
    isw, Jun 8, 2011
    #5
  6. Ted Shoemaker

    Martin Brown Guest

    Actually no it isn't. The sensor originates 10MB of raw independent
    sensor data - compressed at maximum quality that translates to an upper
    bound of about 5MB JPEG if you have the right sort of difficult to
    compress subject material and a *very* steady hand or use a tripod.
    Clear sky is easy to compress so any sky drops the upper limit.

    Even the slightest weakness in your photographic technique and the JPEG
    encoding at top quality is no longer the limiting factor (at least as
    far as resolution is concerned, dynamic range can be an issue at 8bits).
    You are guessing wildly and have no clue what you are talking about.
    Buy a test chart and photograph it under controlled conditions.
    DPreview and plenty of others. AP do resoution test charts. I don't
    think it is much use but YMMV. It is difficult to avoid finding sample
    test shots off different cameras at review sites these days.
    You will have to define what you mean by true resolution. Canons by
    default tend to originate images with slightly soft anti-aliassing you
    can always apply some unsharp masking afterwards according to taste.

    The thing should be able to match its nominal 10Mpixel spec on all but
    the most demanding pathological test cards.

    Regards,
    Martin Brown
     
    Martin Brown, Jun 8, 2011
    #6
  7. Correct! I don't know this stuff. That's why I'm asking.

    Most of your comments were helpful. This one was not.

    Ted Shoemaker
     
    Ted Shoemaker, Jun 8, 2011
    #7
  8. Ted Shoemaker

    Ofnuts Guest

    My 12MPix Canon DSLR produces JPEGs that are usually in the 4.5-5.0MB
    range with the highest quality setting, so the files sizes from your
    camera are fairly typical.

    I suspect you went pixel-peeping the images and found them not so great
    when looked at 100% zoom. This is normal with compact cameras,
    especially when used at high ISOs. I tend to consider that to get from a
    compact camera an image as clean as what I would get from my DSLR, I
    have to average the pixels on 2x2 or even 3x3 squares, so the perceived
    resolution of the compact camera is indeed between 1/4 to 1/10th of the
    announced figures.
     
    Ofnuts, Jun 8, 2011
    #8
  9. Ted Shoemaker

    Alan Browne Guest

    No. If you take a photo of an evenly lit white wall, it will compress
    it even further down. The resolution doesn't change (the amount of
    different information is much less).

    Or take a photo of a highly detailed scene. The JPG will be larger.

    The resolution doesn't change.

    ergo: JPG size has absolutely nothing to do with resolution. The JPG is
    smaller due to near lossless data compression.

    Further, the sensor on that camera is tiny. THAT will affect the actual
    resolution due to noise contamination, esp. when coupled to the tiny
    lens on the front of the camera.
    You can calculate its sampling resolution (pixels in each dimension on
    the sensor). That's much easier.
    dpr.com. All sorts of technical information that misleads people.

    http://www.dpreview.com/products/canon/compacts/canon_a490

    And guess what? The resolution is truly 3648 x 2736 or 9.98 Mpix. Just
    as advertised.
    If it records raw you will get more _depth_ of information (or at least
    less numeric artifacts in it). But you'd never know that from most
    casual photos.

    But frankly, that camera is so limited in control, about the best thing
    you can do to improve the resolution (or sharpness which is probably
    what you really want) is to use a tripod with it - or at least scene
    modes for shallow depth of field where the aperture will be wide open
    and the shutter speed faster. Couple that to ISO speeds of about 400 -
    800 to get even faster shutter speeds.
     
    Alan Browne, Jun 8, 2011
    #9
  10. Ted Shoemaker

    bugbear Guest

    Take it as a polite hint to ask more, and guess less.

    BugBear
     
    bugbear, Jun 9, 2011
    #10
  11. Ted Shoemaker

    Alfred Molon Guest

    Best would be a Siemens star resolution chart, because you notice
    immediately if the camera software is trying to generate fake resolution
    by extending the lines.
     
    Alfred Molon, Jun 9, 2011
    #11
  12. Ted Shoemaker

    Martin Brown Guest

    It may not be particularly helpful to you as the OP but it is immensely
    useful to anyone who reads this thread in the future and might
    mistakenly think that you know what you are talking about.

    Regards,
    Martin Brown
     
    Martin Brown, Jun 9, 2011
    #12
  13. Your camera doesn't interpolate. (Unless you use digital zoom,
    if it has that.)

    mp3 is also a lot of compression.

    [...]
    Which is to be expected.
    You can improve that with unsharp mask.
    And tripod.
    And low ISO.
    (unfortunately the latter 2 don't help much with moving subjects.)

    The relevant measurement is the output medium at it's size
    (i.e. if you print on 4x6, look on 4x6, if you show them on the
    web at 800x600, compare them at 800x600, if you sometimes make
    meter high posters, examine the qualitry on meter high posters).
    That's noise.
    Wrong. You'll still have almost all of the resolution despite the noise.
    It might not look as fine in large print at close distances, though.

    But I can tell you that images shot at an effective ISO of 6400 with a 20D
    (8 Mpix, 1.6x crop) still look OK to good on 28x42cm enlargements with
    minimal postprocessing/noise supression. Even though they look ... ah, not
    so good on the web.

    -Wolfgang
     
    Wolfgang Weisselberg, Jun 9, 2011
    #13
  14. Any Bayer-type camera produces output images where the pixels do not
    correspond
    directly to sensor inputs. I actually think it would be fair to say
    that every single
    pixel was produced by interpolation (from multiple sources).
     
    David Dyer-Bennet, Jun 10, 2011
    #14
  15. Ted Shoemaker

    Guest Guest

    they do spatially.
    foveon also interpolates for every pixel, among the three layers.
     
    Guest, Jun 10, 2011
    #15
  16. [reformatted for long-short-long-short line length]
    That doesn't affect the resolution in any important way,
    however --- unlike, say, digital zoom. Or blowing up the
    image size.

    -Wolfgang
     
    Wolfgang Weisselberg, Jun 11, 2011
    #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.