The digital zoom myth busted

Discussion in 'Digital Cameras' started by bob, Oct 27, 2004.

  1. bob

    bob Guest

    The digital zoom test.

    The equipment included a Nikon Coolpix 5000 camera with Firmware 1.7. I
    used a triopd. I set the camera to manual mode, and made all the exposures
    on the same setting. I used "fine" full resolution .jpg files, because that
    is the setting I most frequently use. I set the camera to 80mm mode (full
    optical telephoto) and made several exposures. Then I used the 4x digital
    zoom and made several exposures.

    After downloading the files to the PC, I used Photoshop 6 to resample the
    optical zoom to 4x as large. That's from 2560x1920 to 10240x7680 with the
    bicubic setting. I cropped the large file to approximately the size of the
    small file. At this point I have two images on the monitor. They have about
    the same number of pixels and about the same field of view.

    Here are the two images, side by side (digital zoom on the right) at 100%:

    http://www.2fiddles.com/photos/UnalteredSampleCopy.tif (273k)

    After applying levels of 36, 0.81, and 178, I arrived at this:

    http://www.2fiddles.com/photos/levelsCopy.tif (351k)

    I provide the first file in case anyone would like to perform experiments
    on the unaltered files. Although I call the optical zoom image
    "unaltered," it really has been altered. I used photoshop to resample it.
    For those who might like to try other resampling software, there is the
    original section of the optical zoom:

    http://www.2fiddles.com/photos/OpticalSample.tif (51k)

    To my eyes, the digital zoom looks better. There is better gradiation in
    the colors. In particular, I'm comparing the green leaf and the orange leaf
    in the lower right hand quadrant. In exchange, the resampled optical zoom
    seems to have a tiny bit more shadow detail, as seen in the pine needles in
    the upper left quadrant.

    The one thing they are not, is the same, which is the myth. I performed
    this test to confirm observations I made under less controlled
    circumstances, where I though the digital zoom rendered tones smoother, but
    the exposures were different, which prevented a good comparison.

    Bob

    PS the samples are all lwz compression PC byte order Photoshop tifs. The
    quicktime plugin does a good job of opening them, but they won't open in at
    least some browsers in "standard trim."
     
    bob, Oct 27, 2004
    #1
    1. Advertisements

  2. My understanding is that the camera will crop and enlarge the area under
    digital zoom *before* conversion to jpg, which would explain differences
    in picture quality.

    Try the experiment again with the camera set to produce RAW or TIF or BMP
    files.
     
    Jacobe Hazzard, Oct 27, 2004
    #2
    1. Advertisements

  3. bob

    usenet Guest

    Kibo informs me that "Jacobe Hazzard"
    That would be my explanation too.
    Betcha digital-zoom isn't available in RAW mode.
     
    usenet, Oct 27, 2004
    #3
  4. wrote in 4ax.com:
    Reminds me of a joke about a heavy metal group band member that was having
    an animated argument with his manager:

    "So let me see if I got this straight. We got live chickens squawking on
    the stage, and we kill the chickens an axe... Then we chop up one of the
    speakers, and we totally trash all the guitars and the drum set, all in
    order to get the most out of the concert, and you decided to record it
    digitally to get better sound ???"
     
    Lasse Vågsæther Karlsen, Oct 27, 2004
    #4
  5. I agree - although my argument /for/ using digital zoom in the past has
    been based on the premise that (providing you stick to 2:1 zoom - perhaps
    4:1 but I've not tested that) the interpolated image in the camera will
    suffer less from the JPEG compression than the full resolution image.

    One added factor in Bob's tests is that he is also "testing" the
    interpolation algorithms used on Photoshop 6.

    Thanks for sharing your results, Bob.

    David
     
    David J Taylor, Oct 27, 2004
    #5
  6. bob

    bob Guest

    It was an interesting experiment, and you're welcome.

    I did consider testing RAW mode, but I do not frequently use RAW mode. I
    wanted to find out how to maximize the quality of my system, given how I
    typically use it.

    RAW mode takes much longer to write files, they are much larger, and
    given typical lighting, don't seem to produce better prints. With my
    camera. TIF files (on my camera) have all the disadvantages of RAW, only
    more so, with none of the advantages.

    I got the camera for it's wide angle abilities and I will continue to use
    it that way. If I found I frequently needed longer focal lengths, I would
    consider the teleconverter, but for now, if I know I'm going to be
    cropping before printing, there seems to be an advantage to using the
    digital zoom, especially where the subject is characterized by smooth
    tones.

    Bob
     
    bob, Oct 27, 2004
    #6
  7. bob

    Owamanga Guest

    He isn't claiming that. Obviously optical zoom is going to be better -
    always. He is simply comparing in-camera digital-zoom with a digital
    zoom that is done afterwards in software by cropping a section from a
    non-digital-zoomed image.
    Nikon's firmware has more data at it's disposal - it has access to the
    12bit raw image which is quite obviously better than attempting to do
    the same from an already compressed 8-bit jpeg outside the camera.
    Now, if he were playing with NEF (RAW) or TIFF instead, the comparison
    may show neither method is superior.

    Here is a test: take a 1/4 section of any picture and compare it's
    histogram to that of the whole picture, they are different right? And
    that's part of the problem with an after-the-jpeg digital-zoom, the
    jpeg was optimized based on the whole scene's levels, not just on the
    section you want to zoom to.

    Now, I must be alone here, but I think the image on the left
    (photoshop's resize) is better than the digital-zoom - much less
    artifacts, and the color difference isn't that big.
    I don't think it's anything special.
     
    Owamanga, Oct 27, 2004
    #7
  8. Jim Townsend wrote:
    []
    Nikon seem to have very good firmware - their JPEGs seem to be able to
    retain picture information in smaller files than other cameras need. I
    suspect the parameters for the compression have been carefully chosen to
    match the lens and sensor characteristics.

    Careful attention to detail.

    Cheers,
    David
     
    David J Taylor, Oct 27, 2004
    #8
  9. bob

    Alan Meyer Guest

    An interesting experiment. Thanks for posting it.

    In theory, as others have pointed out, the camera might be
    able to do a better job than the program because it has access
    to information that is not available to the program. This
    information includes:

    1. More bits of color per pixel.
    2. Unprocessed/unfiltered input.
    3. Data lost by JPEG compression/expansion.

    Number 2 above refers to raw sensor output, before
    amplification and various types of filtering.

    However whether any of this extra information can be used
    to create a perceptibly better interpolation is an open
    question. More information doesn't help at all if it
    turns out to be sensor artifacts and imperceptible
    color distinctions.

    I don't know that the algorithms suitable for interpolated
    images would be different from the ones suitable for
    non-interpolated images, or that the extra bits and slight
    distinctions of color make any difference at all in
    perceptible quality.

    On the other side of the equation the program has
    more processing power and memory at its disposal
    and might possibly use more sophisticated algorithms
    that the camera designers decided would take too much
    processing time or memory on the camera.

    Another issue is that different types images might appear
    differently than your test results. Differences in the amount
    of color, detail, brightness, and other factors could show
    up differently in the two types of interpolation.

    The images looked close enough in quality to me that
    I would base my decision on convenience and personal
    working style.

    Taking a shot with digital zoom allows you to print the
    image without cropping, and to frame the final image
    in your viewfinder or LCD.

    Taking a shot without digital zoom allows you to look
    at the shot differently after you've got it in your computer
    and crop it differently from what you originally had in mind.

    I lean more towards the latter - no digital zoom, but I
    can see why some people might prefer to use it.

    Alan
     
    Alan Meyer, Oct 27, 2004
    #9
  10. bob

    bob Guest

    In fact, both images are processed through Nikon's firmware. Without
    shooting in RAW mode, that is unavoidable. I provided a sample of the
    lesser processed image, in case anyone had access to different software, or
    thought of a beter Photoshop process (or even a newer version) and wanted
    to give it a try.

    Bob
     
    bob, Oct 28, 2004
    #10
  11. bob

    bob Guest

    I'm not so sure about the artifacts. I read most of the dark splotchy stuff
    on the left of each image as artifacts. I kind of think that for any given
    scene, if there is a lot of busy detail, the photoshop resample might
    appear sharper, but if there are large gentle transitions, the nikon
    resample will be smoother.
     
    bob, Oct 28, 2004
    #11
  12. bob

    bob Guest

    That is my main conclusion too. There does not appear to be an obvious
    reason to exclude either method. Perhaps with more testing, I will discover
    a rule of thumb that for some types of images it is always better to use
    method X.

    4x digital zoom is pretty radical too. That takes the 80mm lens up to
    360mm. I bet at lesser digital zoom, the differences between the images is
    less.

    I think the digital zoom has an advantage if you use any automatic meter
    mode -- at least on my camera. The metering is based on the zoomed field of
    view, instead of the entire sensor area. Another slight advantage, if you
    are shooting many photos and don't always clearly remember things, is that
    when you are previewing photos you don't need to remember that you intended
    to crop when you shot the original if you crop it in the camera.

    Bob
     
    bob, Oct 28, 2004
    #12
  13. bob

    Ed Ruf Guest

    No quite true, it is apples and oranges, if you normally use matrix
    metering and AF area Auto(defaut is closest subject) . By using digital
    zoom, you have switched the camera into center-weighted metering and center
    focus modes. Since you say you set the exposure and shutter speed manually
    in your experiment, there is still the focus setting change, so the
    experiment isn't as closely controlled as you may have thought. Unless you
    ensured the same AF area settings and did not mention that to us.

    I do agree with the previous statement/conclusion that the jpeg algorithm
    is only using the cropped image info and this may be a benefit depending
    upon the content.
    ________________________________________________________
    Ed Ruf Lifetime AMA# 344007 ()
    http://EdwardGRuf.com
     
    Ed Ruf, Oct 28, 2004
    #13
  14. bob

    bob Guest

    I did have the camera in manual AF area selection; in this case I don't
    think it would have mattered, since the subject matter was about 100 yards
    (90m) away.

    Now that you mention it, I did notice the metering icon had changed from
    Matrix to Center, but it hadn't registered.

    Even for a relatively inexpensive camera -- I paid $450 -- there is a lot
    of complexity. I think my film camera had a total of 3 controls that
    impacted image quality (focus, aperature, and shutter), where this one has
    at least a score. I know it's not a perfect comparison, because film can be
    a big variable.

    Bob
     
    bob, Oct 28, 2004
    #14
    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.