I’ve been doing digital panoramic photography since 2002. While I have used several non-professional digital cameras, I have always mounted them on the same set of equipment: a Manfrotto 3401B tripod (current equivalent is probably their model MT055XPRO3), topped by a Manfrotto 3415 rotating panoramic head (2020 equivalent model 303), which includes a mounting plate for the camera itself.
The pan head is the key element, assuming that whatever tripod is used gives a stable support for it. Not only must the head be able to rotate smoothly in 360 degrees; it must also make it possible to set the nodal point of the camera at the axis of rotation of the pan head to avoid odd parallax effects in the finished panorama. The nodal point of a conventional camera body is not automatically place properly just by mounting the camera on a generic rotating tripod head using the camera’s mounting socket. The pan head I use has two slider elements that allow the mounted camera to be adjusted along two horizontal dimensions. With my older cameras I got good results just by eyeballing the camera above while I adjusted its position to where the obvious axis of rotation of the tripod was vertically line with where I thought the nodal point of the camera was likely be inside its body. Since cellphone cameras are so thin, adjusting one on the pan head would be much easier.
The other necessary feature of the pan head is a way to level the pan head precisely – and certainly not by trying to do that by adjusting the lengths of the tripod legs. The Manfrotto head has a three-point adjustment system with a bubble housed under a target window. Poor leveling leads to panoramas that a vertically shorter than they might otherwise be, since the various frames lose data at their tops or bottoms as the assembly program fails to find pixel matches there.
The Manfrotto head also has an indexing system that allows the head and camera to be rotated around the vertical axis in precise degree increments, so that at the individual frames will overlap consistently. The choices range from 10-degree increments (36 frames) to 90-degree increments (4 frames). I almost always use the 24-frame choice (15-degree increments), sometimes the 20-frame choice (18 degrees). I tried the 12-frame setting a couple times but worried that the frames would not overlap enough. For whatever reason, I always rotate the head clockwise (as seen from above), and with a conventional digital camera used a cable release to avoid nudging the camera out of position.
After that it was a matter of chanting to myself “shot 1, click, pause for storage, move camera” and on and on. I almost always take at least two sets of exposures for each panorama, and sometimes more sets or just more single shots where an undesired incidental element is in the camera, especially something or someone that is moving. That helps avoid having parts of a person or thing (car, especially) disturbingly absent from the finished panorama.
I suppose that what has changed least in the years I have been doing panoramic photography is, at least at the medium-price ranger, the mechanical elements – that is, the tripod and, even, the pan head. However, as everyone who takes pictures would know, the photographic process has changed radically, from primarily film to primarily digital. After digital photography processing software became available, there certainly were people who took their panorama frames on film and then scanned them so they could be “stitched” into panoramas.
My earliest panoramas began with a Nikon E800, which made exposures of 1600 x 1200 pixels, producing a file size of about 650 KB. In 2006 I was using a Panasonic DMC-LZ2 camera, which yielded a picture size of 1920 x 2560 pixels, with a file size of 2.3 MB. In 2014 I used a Canon PowerShot SD1300 IS to produce frames of 3000 x 4000 pixels with a file size of 3.4 MB. This level of quality is, of course, much below what can be achieved with professional-level equipment now (late 2020), which works at 12M. I suppose picture size (dimensions, that is) is irrelevant, since pictures can be enlarged or reduced digitally. But the raw file size does give a good idea of the pictorial quality one begins with for “stitching” and other panoramic processing. And for more comparison: professional digital cameras of 2020 offer 45MB. Since I am not a trained photographer I can’t offer comparisons of any real sophistication. But those are readily available from specialized sources.
In my experience, at least given the time I began making panoramas and the equipment I have used, the greatest change in the technique has been the software that stitches the frames into a single 360-degree picture file, wraps and bends that picture into a virtual spherical perspective, and then allows that file to be viewed on a computer, tablet, or mobile screen, usually with various accessory facilities like links to other panoramas, other media, or external websites, and addition of sound. The first software I used was Apple’s QuickTime Virtual Reality (QTVR) program. That technique is explained on the page about “Modern Panoramic Photography”, reached from the main link to “Panoramic Photography” above.
I used a few of QTVR’s accessory features, like hot spots to static views associated with the panorama site. While my first panoramas were of sites of my private interest, such as vacation spots, I soon began to think about using QTVR in my profession of language teaching and thus making it part of my website and also the learning software I was creating. The link “My Panoramas Made with Apple QTVR” leads to examples of those panoramas, including some that were made from the same frames used to make the panoramas offered by my main gallery. The difference in quality is due far less to the difference between, on the one hand, QTVR and, on the other hand, PTGui and Pano2VR, my current software, than to an increase in my abilities to improve my frames in ordinary graphics-processing software.
Early on, though I had available good graphics software (Photoshop), I knew only how to adjust brightness and contrast, to copy and paste picture parts between repeat exposures of the same view in order to remove unwanted incidental elements, such as moving cars or unsightly creatures, and to retouch crudely with paint tools and such. That remained my level of quality as I tried, after QTVR was discontinued, to use other software (VR-Worx, etc.) to produce panoramas that still used the .mov format. When the QTVR file format, .mov, became non-standard in media distribution I turned my attention from panoramic photography to other quite unrelated projects.
After I retired (2015/16) I again looked into panoramic software, with the motivation, I admit, of seeing whether those panoramas and later ones could be part of whatever my legacy might be. I was also thinking about finding people with whom I shared interests. The big decision, after I researched the stitching and presenting software and found no way just to convert the .mov QTVR files to something else, was accepting that I would have to begin anew with the original sets of raw exposures, and that the method would produce panoramas that would not ever exist as single files to be viewed with some sort of player, but rather collections of “tiles” that would be download as needed by the browser as the person viewing the panorama navigated within it. Yes, the Garden Gnome “package” method does allow for viewing an entire collection of tiles that can be uploaded and downloaded as a unit, but that requires that the viewer get the Garden Gnome Package Viewer and deal with a large one-chunk download. On the whole, I am pleased with the shift from single-chunk .mov-type files to the “tiles” method, and of course Pano2VR, even the “lite” version, offers presentational tools that are much more sophisticated than Apple QTVR did – or ever could have done, considering how far back in time that was. The specifics of Pano2VR are discussed on the page about “Modern Panoramic Photography”.
My graphics processing skills remain primitive by professional standards. First, I use the free package GIMP rather than Photoshop with its pay-for-rental policy. Still, GIMP offers much more than I can handle even now, and the improving images with it has produced manifestly better panoramas. Four procedures have been of immense value to me:
- When I want to remove an undesired object or creature, or a perhaps one that is desirable but is nevertheless being dismembered in the stitching process, I select from whatever duplicates I have of the target frame any exposures that lack the undesirable element. Rather than, as before, copying and pasting from one image to another, I stack the “clean” version beneath the imperfect one, adjust for the minor movement of the camera between shots, create a layer mask with an alpha (transparency) channel, and then, on the top layer, remove the undesirable content and let the good part of the lower image show through the mask, with some blurring of the border of the zone to obscure any lines caused by slight differences in exposure or alignment. Sometimes two exposures of the same view will each have an undesirable element that can be replaced by a “clean” area on the other. That requires stacking of one or more copies of each exposure.
- To lighten dark areas and darken over-light areas I always begin by creating a negative overlay from a copy of the original. This is placed above the original. Initially the combination of the two looks awful (blurry, blotchy), but that is cured by applying a Gaussian blur of, usually, between 100 and, in rare instances, 1000.
- Sometimes it helps further here to add another layer. This is done by copying the original, adding to the new layer a mask (grayscale, with alpha channel), then inverting (menu:color) and, finally, setting the mode to “screen”.
I always apply the mask/overlay techniques to the stitched panorama, and sometimes use them as well when a set of exposures contains some shots that are much underexposed and others, often adjacent, that are much overexposed. This can happen when the camera points at, say, a building, especially close up, and rotates to point away from the dark area to an area that is well lit.
4. Sometimes a sky, in a single frame or an entire stitched panorama file, is so blown out that it is necessary to add a fake sky, often a blue one since moderately blown out skies on overcast days can resemble gray ones. The procedures involves three steps: a) creating a new layer that uses the gradient tool and the color palette to generate a graphic that extends from darker blue at the top to lighter blue lower down; b) adding a layer mask and transparency channel to the original picture, quickly selecting (rectangle, lasso) and removing large areas of the sky and then, much more laborious, using the eraser and fuzzy-select tools to trim away the bad sky to the edges of whatever is obscuring the sky. The fake sky, on its layer below the main pictures with its mask, will then show through. When trimming away the skyline to allow the fake sky to replace the blown-out real sky, it is better to remove slightly too much of the (unblown-out) skyline than too little. Failing to do that will leave a narrow but very conspicuous bright line of leftover blown-out sky between the skyline and the new (and darker) fake sky. Even when the panorama is zoomed-in on, it is very difficult to notice that part of a distance tree or rock formation has been cut off, especially if the eraser is set to partial opacity. That produces a gradual zone between the center of the eraser and its edges.
A few times I have had only a single set of exposures to work with for a given poorly exposed frame that has not only a blown-out sky but other elements that are blown out too. The technique I describe here will work only with distant landscapes and very distant architectural elements. A good bad example is a shot in the Alps were not only the sky but mountains that have both snow and bare rock are blown out or very light. If the panorama set is taken with many shots and thus a relatively small angle between adjacent views, and if an element that is blown out in one picture appears fairly solid in an adjacent one, it is possible to put the good picture beneath the bad one, roughly line up the target pairing areas, add a layer mask with an alpha channel to the defective image, and then us the eraser and other tools to let the better view show through the bad one.
That may itself be insufficient, or it may be advisable to use distortion tools (sheer, etc.) to pull the good element into line with the shape of the bad area. It may also help to make a temporary copy of the bad picture and, by changing color and contrast, create a darker version of the parts of the target image that are not completely blown out. When this layer is put beneath the masked target image and the latter’s opacity is adjusted downward, the darkened area will help guide the digital tools as they remove the blown-out areas. What is necessary here is to have the erasures active on the target picture while looking at the helper layer.
If I take many panorama sets at nearby camera locations, in order, for example, to show different parts of a town that share the same approximate background, such as distant mountains, I can steal a clear background from a frame in one set of images to replace a blown-out background in another. Sometimes I have to use the sheer tool and such to adjust the background to fit its new place in the other set of images. But here, and certainly when the background is replaced by part of an adjacent image in the same set, the stitching software on its own can make a lot of adjustments, since its job anyway is to bend the edge zones of one picture and those of its adjacent partners to blend together. Slight retouching may be necessary, and since this is a background and often distant, the rubber-stamp (clone) tool can do wonders, such as with repairing odd spots in forests or rock formations.
Here are part of a single frame and the PTGui build for the entire panorama (much reduced) in their “before” and “after” appearances, and then the resulting panorama itself:
And here is the finished panorama:
I think the most dramatic improvement in a set of photos I worked with was a nighttime view taken in Venice from the Fondamenta San Giovanni on the island of Giudecca. Here are the original PTGui build, its improvement in GIMP, and the resulting Pano2VR panorama:
PTGui works well automatically to stitch most sets of panorama pictures that are taken with reasonably correct exposure on a stable tripod with consistent degrees of rotation. I suspect that even a homemade angle indicator would work OK with a conventional rotating camera mount as an indexing device. But one night in Venice I took a set of images that contained many exposures that were approximately satisfactory when they included nearby features, but were otherwise so dark, as with the ones taken over water to distant features, even the center of Venice, where nothing appeared except a few dim lights on the horizon. I worked long and hard to help PTGui generate control points by hand, but nothing worked to produce a usable stitch, not even when I temporarily enhanced the exposures to bring out whatever might have been recorded, even faintly. Oddly enough, though, back in 2006(!) the QuickTimeVR software had, even without visual enhancement or other aid from me, produced a viewable panorama from that same set of pictures. Unfortunately, back then I did not know how to improve the raw images for even a decent set of exposures, and now I don’t have a working version of QuickTimeVR. That particular panorama is available on my gallery of .mov panoramas, but you have to have a way to view a .mov file that is not a conventional movie clip but rather a panorama.
So, at least in a few instances, newer is not necessarily better. But I happily give up on that one panorama for what I have gained with PTGui, GIMP, and Pano2VR.
So far I have been able to construct new and better panoramas from my sets of original individual frames. For years I tried to find a way to salvage a .mov QTVR panorama whose original photos I could not find. It was near and dear to me: my first panorama ever, and of Waldo Lake in Oregon. Then I read a post by a certain “johnh”:
http://www.tawbaware.com/forum2/viewtopic.php?t=2199
Johnh – many thanks! – recommended taking about 8 screen shots of the old panorama on a full-screen display and then stitching those new-old pics. To save as much data at the top and bottom in the stitching process, and to ensure a good overlap, I took 26 screen shots, where I had originally taken 24 exposures with camera and panoramic tripod head. The unusual number is because I used the “play” arrow at the center of the QuickTime 7 Player control panel and then, for each shot, noted a low-down distinctive image element that I could use as an index point by moving rotating the panorama until that point was over the “play” arrow. The procedure worked far better than I had expected: maybe 1-2% of the original height lost in the newly stitched rectilinear panorama at bottom and – just sky anyway – top, and the stitching went with no need to add control points manually. Yes, the resolution is lower than on my later panoramas, and also lower than that of the original QTVR panoramas. But better that than nothing. Here are a screen shot from the old Apple QTVR .mov panorama, the new Pano2VR panorama made from the screen shots, and – for those who have the old QuickTIme viewer that will display QTVR files in the .mov format, the old panorama. IMPORTANT: Rather than trying to play the QTVR .mov file directly, download the file (12.7 MB) to your computer and then open it with the heritage version of QuickTime (probably v. 7).
