Digital Mosaics: Surpassing Large Format Film Images

Discussion in 'Digital Photography' started by Roger N. Clark (change username to rnclark), Aug 26, 2006.

  1. Hi.
    I have been doing large format photography for going on 2 decades.
    I have also been doing wildlife photography, first with 35mm then
    digital. Thus, I often carry both digital wildlife and 4x5 gear
    on a hike (up to 70 pounds). That gets real tiring and limits
    my range (and, obviously, I'm getting older). I want an alternative
    without giving up anything ;-). I switched to a Toho 4x5 field camera
    (3 pounds) from heavier cameras several years ago, but it is still
    too much weight doing both wildlife with digital and scenics with
    4x5.

    Mosaicking many digital image frames has intrigued me for
    some time, and I have been experimenting with the methods,
    from field to computer processing. Like large format view cameras
    and methods, there is much to learn. But my experience so far
    is that digital mosaics can equal and surpass 4x5 drum scanned
    film in many applications, including large depth of field imaging
    requiring tilts on a view camera. And I can get images in the
    field faster and under conditions not suitable for large format
    photography (like wind).

    I've written up some of my experiences in this article and compare
    the digital results to drum scanned 4x5 film images:

    Large Digital Mosaics as a Substitute for Large Format Film
    http://www.clarkvision.com/photoinfo/large_mosaics

    Comments welcome.

    Roger Clark
    my photos at: http://www.clarkvision.com
     
    Roger N. Clark (change username to rnclark), Aug 26, 2006
    #1
    1. Advertisements

  2. Spectacular.

    You mention problems with focus from frame to frame.

    Do you manually focus and lock exposure from frame to frame?
     
    Scott in Florida, Aug 26, 2006
    #2
    1. Advertisements

  3. Thanks Scott.

    I determine the best overall exposure then go to manual and keep
    f/stop and exposure constant for the entire sequence.

    I autofocus separately for each frame, usually using 1 focus
    point. In general I have no problems focusing.
    The issue is: have I set a small enough aperture
    to give enough depth of field so that when focus changes from
    frame to frame, there is not an area in poor focus in the final
    mosaic.

    Roger
     
    Roger N. Clark (change username to rnclark), Aug 26, 2006
    #3
  4. So your focus point can change frame to frame?

    Would it be worthwhile to pick a focus point and go back to manual
    focus?

    Can you notice much change in the images with different focus points?

    The reason I ask, is I am going to do one on a Light House in southern
    Maine (Nubble) my next trip up.
     
    Scott in Florida, Aug 26, 2006
    #4
  5. Roger N. Clark (change username to rnclark)

    Bill Funk Guest

    I just did a presentation on panos/mosaics for a digital photo SIG
    this week, and I used Autostitch.
    I covered setting the aperture for a wide depth of field (if that's
    what's wanted), which seems like a good thing! :)
    I also covered the need to be careful about lens selection. When I
    used my EF 17-40mm lens, since it's rectilinear, Autostitch was able
    to make the resulting pano look good by rotating the different
    individual images to get a non-keystoned final image.
    http://pippina.us/images/glendale 1.jpg
    But, using my Panny FX01, which doesn't have a rectilinear lens, there
    are problems with the pincushioning at wide angles; the obvious
    solution is to use a longer focal length, and take more shots to feed
    to Autostitch.
    http://pippina.us/images/glendale 1.jpg
    :)
     
    Bill Funk, Aug 26, 2006
    #5
  6. Roger N. Clark (change username to rnclark) wrote:
    []
    Many thanks for writing that up, Roger. A fascinating read. You will
    doubtless be looking for ways to speed the post-processing.

    And I thought it was just me who was fussy about weight!

    Cheers,
    David
     
    David J Taylor, Aug 26, 2006
    #6
  7. Roger N. Clark (change username to rnclark)

    ASAAR Guest

    Anxiously awaiting

    "Large Digital Mosaics as a Substitute for Large DSLRs".

    :)
     
    ASAAR, Aug 26, 2006
    #7
  8. I'm waiting to hear how you can do even better it with an array of Fuji
    F30s!

    David
     
    David J Taylor, Aug 26, 2006
    #8
  9. "Roger N. Clark (change username to rnclark)" <>
    wrote in message SNIP
    Congratulations on a successful, and beautiful, stitched image.
    It inspires to experiment more with variable focus distances. The
    potential magnification differences (esp. at shorter distances) can be
    addressed in the stitching optimizer with the additional per image
    optimization of the focal length parameter.

    To address some of the points you mentioned in your write-up:

    - As for the Pros, I agree with the points mentioned, in particular
    about the lower restrictions for DOF, because the full image doesn't
    have to be in acceptable focus, but 'only' the partial frame inside
    the final overlap does. The resulting faster shutterspeed helps in
    reducing subject (wind)motion.

    - As for the Cons, I think the depth of field matching could
    (depending on subject) be significantly helped by using a Tilt and
    Shift lens. I often use the T/S-E 45mm f/2.8 for stitched images. It
    would also allow to reduce the number of images, which benefits the
    amount of post-processing and reduces the risk of changing light
    conditions. Alternatively one could consider an additional program
    like Helicon Focus
    <http://www.heliconfocus.com/pages/focus_overview.html>.

    - The Photoshop layers can potentially be skipped for the most part,
    when you use SmartBlend. It'll adjust for small brightness
    differences, and it does a remarkable job of blending between images
    with movement (ghosting), all on full automatic. The downside is that
    it takes its time doing it, but that's not too much of an issue if you
    let it run in otherwise idle time.
     
    Bart van der Wolf, Aug 26, 2006
    #9
  10. Roger N. Clark (change username to rnclark)

    ASAAR Guest

    That's kinda sleazy, coming as it does from he who finds even more
    ways to push and plug Panasonics than the most rabid Canonistas do
    their precious Canons. FWIW, I've said that those F10/F11/F30
    Fujis, even though they're excellent low light cameras, aren't my
    choice for a number of reasons. Not only that, I've chided kinga
    several times for his stupidly hyperbolic F30 posts. Are you trying
    to sound like a little boy?
     
    ASAAR, Aug 26, 2006
    #10
  11. Been there, done that (with a Powershot G3),
    <http://www.xs4all.nl/~bvdwolf/main/downloads/BerlinerDom.jpg> 6
    images at 1/3rd of original pixel dimensions, and still moved on to a
    DSLR ... ;-)
     
    Bart van der Wolf, Aug 26, 2006
    #11
  12. Yep,
    You can surpass the pixel count easily, but it is harder
    (though not impossible) to surpass the photon count and low noise.
    To surpass the photon count, one would need perfectly still
    scenes, and average several frames for each mosaic position.

    Roger
     
    Roger N. Clark (change username to rnclark), Aug 26, 2006
    #12
  13. If you have adequate depth of field, yes. The advantage of
    changing the focus point is to get good local depth of
    field without having to go to all extremes in the image.
    If your depth of field is adequate, the change is negligible.
    Try it. Use several apertures and see which works best.

    Roger
     
    Roger N. Clark (change username to rnclark), Aug 26, 2006
    #13
  14. Bill,
    Aren't these the same images? Looks like the same link to me.

    Roger
     
    Roger N. Clark (change username to rnclark), Aug 26, 2006
    #14
  15. You could get better photon count on a small-sensor camera by taking
    enough frames to get a mosaic at 2x or 3x the desired resolution, and
    then use binning to get down to the final resolution.

    It would probably make sense to do the binning before the stitching, to
    keep the computational load down.

    -dms
     
    Daniel Silevitch, Aug 26, 2006
    #15
  16. Roger N. Clark (change username to rnclark)

    ASAAR Guest

    I assume that you didn't purchase the DSLR just to do panoramics,
    but took advantage of their superior panoramic advantages, compared
    to large format film. If DSLRs didn't exist, smaller dcams could
    take their place. They do take more time and effort (and frames)
    compared with DSLRs, but they do allow anyone to take panoramics
    even if they don't have sufficient resources to own a DSLR and one
    or more lenses, each one probably more expensive than the P&S.
    While my old Powershot S10 has a panoramic mode, your G3 and any
    recent DSLR should be far superior for that purpose. But there are
    a good number of high resolution non-DSLRs that have excellent
    lenses and that should also be capable of producing very large, high
    quality panoramics nearly as easily as DSLRs.
     
    ASAAR, Aug 26, 2006
    #16
  17. Bart,
    Thanks.

    Yes, I agree. PTGui has advanced modes for working with tilt-shift lenses.
    I do not have one, but I may rethink that. One question is if
    you set the tilt for a local frame in the foreground, what happens
    when you raise up to the horizon where you would not want tilt
    in that frame?
    I'll have to look at that.

    Roger
     
    Roger N. Clark (change username to rnclark), Aug 26, 2006
    #17
  18. Yes, good point. But it becomes difficult rather quickly.
    The 1D Mark II with its 8+ micron pixels holds up to
    80,000 electrons/pixel. A small pixel camera may hold less
    than 15,000, or over 5x less. One would need more than
    5 frames of small sensor size to match the large sensor
    (assuming each sensor had the same pixel count). But that
    doesn't include overlap. A mosaic needs overlap, so instead of
    a 5 array mosaic, you would need to frame more like 8 or 9 to
    match a single large sensor image.

    Then if the large camera mosaic did 50 frames for an image
    (including overlap), the small sensor would need more than
    400 frames to get the same spatial resolution and photon count
    per pixel. But it can be done!

    Roger
     
    Roger N. Clark (change username to rnclark), Aug 26, 2006
    #18
  19. Roger N. Clark (change username to rnclark)

    ASAAR Guest

    But then since the large camera is producing a mosaic, with its
    own overlapped images, instead of the single image you introduced to
    provide a very short lived advantage that never should have been
    made, it quickly reverts back to nearly a 5:1 ratio instead of the
    8:1 or 9:1 ratio you seem to have left for us to consider. I don't
    think anyone suggested that only a single DSLR frame would be used.
    In fact it was your own paper and the title of this thread that
    suggested a DSLR mosaic. Others (and we know who they are) often
    resort to misleading statements and slight of hand to bolster their
    biases. It's disappointing, given your more scientific approach, to
    see you also engaging in suggestive, but misleading "spin",
    especially when it's not needed to prove your point.
     
    ASAAR, Aug 27, 2006
    #19
  20. But you need the same overlap fraction on the DSLR. You're applying the
    overlap fraction for the small-sensor camera twice, which is the
    equivalent of saying "Take a 9-frame mosaic, stitch it and bin it back
    down; now take a bunch of these and stitch them together to make the
    final mosaic". In reality, anybody trying to make a small-sensor
    high-photon-count image would do it in a single pass, but with smaller
    angular coverage for each frame than the DSLR user would use.

    So, it would be 250ish frames for the small camera, not 400ish.

    -dms
     
    Daniel Silevitch, Aug 27, 2006
    #20
    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.