QTVR panoramas

Enormously high-res QTVR panorama of Union Square (from here.)

Enormously high-res QTVR panorama of Oxford University Museum (from here.)

The Oxford panorama was made by stitching 9 images together, which explains the resolution. It seems that the Union Square panorama was done with a 10.5mm lens, which I think is around 180° diagonally, so that must be multiple stitched shots as well.

These really put my attempts at QTVR panoramas to shame. I guess it's just not possible to get high resolution panoramas out of a single shot: you need multiple images' worth of pixels to get the resolution up.

But resolution aside, I still can't figure out how to get shots with this Kaidan lens that have enough depth of field to actually be in focus. It's very frustrating, especially since I can't even really tell whether it's in focus by looking in the camera: I don't know whether it worked until I get home and run the processing software.

I'm not sure if I'm doing something wrong, or if it's just junk.

Tags: , ,

15 Responses:

  1. ashura93 says:

    The Oxford one doesn't even look like it's HDR / tone mapped... but i presume it must be because the exposure is pretty close to perfect everywhere.

    Time to get your hate on... although not directly related to your Kaidan lens, this Panotools tutorial covers combining multiple shots to get the exposure right in a spherical panorama.

    I'd guess that you could do roughly the same thing with your Kaidan lens, take multiple exposures and tone map.

  2. curlyeric says:

    DOF/Focus is easy enough. You need to use manual or aperture priority mode on you camera. Set the aperture as large (number wise) as it goes and use a tripod if that causes your shutter speed to get too slow. That will ensure a sharp image for objects off all distance.

    • curlyeric says:

      If you have access to Photoshop CS3 they have a new feature where you can stitch together many images. That might help with the exposure and resolution issues as well.

  3. edouardp says:

    Those high res panaramas look great - I've been following the work that appears at http://www.panoramas.dk/ for several years now, and some of their shots are stunning too.

    You're probably right about the number of pixels with your Kaiden lens - it's appealing because it's simple and only requires a single shot, but after you do the maths, there just aren't enough pixels to come up with a very high quality image. Which is a shame, because I thought it was such a cool idea when I first saw them on sale...

    As for the focus, that sucks - it's simple enough optics, so it must surely be possible to get it right? Or perhaps it's a tolerances thing, or maybe the lens rectilinear correction is only approximate? No real idea, sorry.

  4. Nice panoramas, but your union square sucks. Sorry.

  5. cylan says:

    I think that you've got it when you say that multi-image stitched panoramas will always be superior in resolution than a single image from a parabolic mirror, simply due to the sheer difference in the amount of information captured.

    However, you're right - the images you seem to be producing from the mirror attachment are very unsharp. I don't know what camera and lens you're using with your kit, but somewhere I think you mentioned that you've been focusing the camera at infinity when you take your shots. You should actually be doing the opposite and focusing close - that is, to get "infinity" in focus you need to focus at the focal point of the reflector, or on the virtual image projected by the reflector, which is much closer than infinity. To be in the ballpark, you need to focus at a distance that comes close to the distance from your lens to the reflector's surface.

    I don't actually have one of these to play with, so I can't be 100% sure of this. But take a look at the last diagram in the illustration on the Wikipedia entry for focal length - the F is where you want to set your focus to. Also, the instructions for the device seem to indicate that a macro lens is best suited for using the reflector, which indicates to me that you're meant to focus close, not far.

    Also, try using your eyes in an experiment - take the reflector (or the back of a spoon, etc.), hold it close your eye (say no more than a foot away), and position yourself so that something you can see in the reflection is about as far away as something else that is otherwise in your direct line of sight (these should be maybe three feet or more away). Now look into the reflection - your eye will have to focus close to see the reflection clearly. Now look away from the reflection to the other object, and your eye will have to refocus, even though the object seen in the reflection should be about the same distance away. If focus distance for infinity through the lens = focus distance for infinity against the reflection, then your eye wouldn't have to do this. But it does, because you're actually focusing on the mirror's virtual image, which is very close in comparison to how far away the objects really are.

    (If the reflector were a fully standalone "lens", then you would focus at "infinity" to focus in infinity, but this device actually acts more like a screen that warps a large area of space onto itself. Also, if it were a flat mirror you'd focus at infinity to get infinity too)

    Hope you find this helpful (and that I'm not totally off-base here).

    • jwz says:

      Yeah, I'm not actually focusing at infinity - the picture is obviously blurry when you do that. It's not clear to me whether I'm focusing on the surface of the mirror, or "past" it somehow, but what I've been doing is focusing manually and eyeballing it until it looks like the mid-range objects are in focus. But, that's really hard to do with a digital SLR, since they don't have those cute focus-prisms that film SLRs did. So, I've been focusing on something, and then hoping the depth of field is large enough that the rest will be in focus too... but mostly, it's not.

      I'm using a macro lens (Sigma 50mm F2.8 DG) which is one of the lenses they recommended on their site. (As expected, the mirror attachment doesn't work at all with my non-macro 50mm.)

      So, presumably the real problem here is that I have insufficient depth of field, which means that even F9 is too low. But going higher than that means taking, like, ten second exposures.

      • cetan says:

        I doubt it would be 10 seconds, but this is clearly work for a good tripod rather than our shaky arms. Either that or bump up the ISO to 1600 and see if you can get shutter speeds you can hand-hold at.

        Stopping down to f/16 would be a good idea. Anything past that and you're going to soften the image due to diffraction effects.

        • edouardp says:

          Diffraction effects beyond F16? Practically I don't think so.

          I've always enjoyed taking very deep focus pictures with wide angle lenses at F22 (since that the smallest aperture I've got), and never noticed degradation of the picture. I understand how diffraction works, and that it increases as the diameter of the hole the light passes through decreases in size. I even understand airy disk calculations.

          But, even top end SLRs have all sorts of other limiting factors that generally overwhelm the diffraction limitation. Digital cameras even more so (unless you have a camera with one of those Foveon sensors - all other digital cameras are half, or less, the megapixel resolution when it comes to making calculations like the ones needed for diffraction limits).

            • edouardp says:

              Yes - I see, from those examples, that f22 is better than f11.

              I really do understand the limitations in various directions (lens aberration at f5.6 with a so-so lens is shockingly bad), but unless you are taking photos of lines, the diffraction effects at f22 are minor enough that you shouldn't worry about them, no matter what the maths says.

              And I think, for Jamie's focus problem, the extra f-stop would help more than the very minor softening he'd be getting.

  6. neontotem says:

    Tangentally related to the cubic cutout: Eddie Elliot's Video Streamer. You may recall his work was featured in the premiere edition of Wired magazine. Main premise: Take frames from video and stack them sequentially ala a box of business cards. Edge pixels reveal patterns. Use of a clipping plane reveals patterns. He also offers the cubic cutout treatment of a few videos on his site.

    Always thought a mash-up between this and Dan Kaminsky's Angel Project would yield interesting results.