Over the last couple of months, I’ve spent some time working with a 4x5 studio camera and medium format digital back (although you could also make this process work with a mirrorless camera) to create stitched photos of still scenes that are close to, if not quite the size of large format photos (I usually max out around 3.5"x4.5"). I wrote a series of blog entries on the process, in case anyone else is interested in giving this (admittedly, pretty niche) technique a try.
Part I outlines the general theory behind the technique.
Part II goes over the details of actually capturing digital images on a view camera, from a hardware perspective.
Part III gets into how you set up and shoot the scene, and
Part IV covers the process of stitching the constituent pieces into a final image.
Well…looking at them on a screen, so far :p. In theory I could use them to make some enormous prints, but I’ve never had any problems making wall prints from my medium format (or even m4/3) single exposures, so they’re not really necessary for that. Mostly what I’m looking for here is the larger sensor look and the ability to use view camera movements, the enormous resolution is just a bonus (although it’s also a bit of a pain when it comes to storing the damn things).
It’s also kind of neat that with a monorail camera you can pretty much get as much macro magnification as you want by just adding more and more extension rails (I can get upwards of 2:1 magnification out of a 150mm lens with the 24" of rails I have now, and you can always add more), although I haven’t really done a ton of that yet. I’ll probably play around with it some more in the future, and the nice thing is that I’ve found when doing macro I can use camera movements to help bring more of the image into focus in a single frame, whereas with parallel image and lens planes you’d need to use a crazy small aperture and/or focus stacking. Of course, the downside is that when your image size is getting up close to 4x5, even 2:1 magnification isn’t going to be particularly impressive.
Sorry, but I did not quite understand your technique.
Would it be possible to get almost the same result by moving the camera and taking pictures with the camera’s distance and position constant. Then combine the obtained flat pictures with Hugin.
The trick is that if you move the camera, then the camera’s position isn’t constant ;). For instance, imagine your subject is two objects of the same size and shape placed directly in front of you such that the nearer one eclipses the farther.
Let’s say the objects are spheres one inch across. If you now move the camera an inch to the left, your scene has completely changed. Instead of seeing just one sphere, you now see the second peeking out from behind the first. Unless everything in your scene is very very far away, that kind of parallax will make it impossible to stitch the images together in any kind of a coherent way
@bieber Ok, you are right. Moving the camera is successful when the subject is flat like a poster. I have used this technique when I photographed graffiti.
Is the technology you describe possible with a standard 1.6 crop DSLR?
When stitching panoramas you generally have only a few methods at your disposal:
Pan the camera in an orthogonal fashion to the subject.
Shift the sensor relative to the image circle.
Rotate the camera around the subject using the subject as the center of rotation.
Change the yaw and pitch of the camera around the no-/least-parallax point.
Method 1 is uncommon. You would use it when you cannot get far-enough from the subject to use method 4. The huge downside is that it requires that the subject is at a relatively constant distance from the camera, else serious parallax errors will appear. Example: photograph a row of house-fronts on a narrow street, or a plane photographing the earth with the camera pointed straight down.
Method 2 is uncommon. I believe it is what @bieber used. The camera as a whole does not move; the rear standard moves relative to the front standard, so you’re either moving the sensor within the image circle, or you’re moving the image circle around the sensor. It requires that you have a camera with a movable sensor and/or lens, so you cannot do that with a typical DSLR or mirrorless. Shifting the sensor relative to the image circle could be easier than shifting or pivoting the whole camera if it’s a large rig, but the downsides are that you are limited to the area of the image circle, and that the photos will not be using the optically-best area of the lens - the lens is sharpest in the center, but only the center photo uses the center of the image circle.
Method 3 is very unusual, it’s not really panorama stitching, but I thought I’d include it. You would orbit the camera around the subject, constantly pointing at the center of rotation, taking photos every few degrees. It’s not really panorama stitching, since you’d only be aligning the shots, not stitching them but blending them.
Method 4 is most common. It requires that you use a panoramic head to allow rotating around the least-/no-parallax point. It allows you to use the sharpest area of the lens - the center - for each shot, where the overlap determines how much of the peripheral area ends up in the stitched panorama. If your lens is lousy, increase overlap.
There are other possibilities or variations, for example Google Street View basically uses method 4 but using multiple cameras instead of panning and tilting one camera, and smartphones allow you to shoot panoramas also using method 4 except most people don’t pan the phone around the no-parallax point so the results look horrible.
This is awesome. I wonder to what extent this could be done by using a dSLR with like a 50mm lens on instead of a sensor. Unfortunately due to budget constraints I have to think of alternative ways of doing such things. A nice view camera can get expensive, not to mention the sensor
@Morgan_Hardwood: Yep, this is #3. It’s worth noting that while yes, you won’t be capturing the dead center of the lens in each frame, all of your captures are still going to be well within the boundaries of the 4x5 area that the lens is designed to cover (plus most large format lens will cover significantly more than that so there’s room for camera movements), so you’re still effectively capturing the center of the lens’ intended usable area.
@yteaot, @stefan.chirila: I don’t think you could do it with a DSLR on its own, as if you separated the lens from the body it would be too far away for infinity focus (unless of course your goal was close focus). Even then though, you’re going to have trouble mounting the camera and lens separately and parallel to each others planes, and shifting the camera around. Plus the lens wouldn’t have a big enough image circle to accommodate much shift.
You can however mount your camera to the back of a view camera with an adapter like this, and you could probably find an older monorail camera for a few hundred on eBay. Putting a DSLR on the back can be a little problematic though, because the mirror box could end up casting a shadow on the sensor when you use camera movements. From what I’ve heard mirrorless cameras actually work pretty well for this kind of thing though.
