How to properly capture photos for merging?

Question

When photographing photos with the intent to stitch (merge) images into a large panoramic, either manually or with the aid of software, how should one capture the source images for consistency between the frames and to reduce distortion?

Answer 1

I have stitched a few images so far, the successes are all visible on my flickr there are currently 97 there.

All of these were shot off a compact super zoom (Ricoh R7 and CX-1), stitched using the open source Hugin stack and generally with automatic exposure, white balance and focus.

From these experiences I would say that there is really only one thing that matters for creating decent panoramas and that is controlling parallax effects.

It used to be that barrel distortions and varying exposure / white balance wrecked havoc with the stitching process, but newer stitching software can readily deal with these complications. Leaving only the parallax artifacts.

Parallax artifacts are a function of how you rotate the camera while taking the images. If you rotate the camera wrong then the foreground and background won't move together. You can test for this by aligning a foreground and background object, rotating the camera and checking if the two are still aligned, if they are not then you will have parallax problems.

There are two ways to address parallax issues, the first is to take images that consist entirely of distant objects. The second is to take the time and effort to figure out how to rotate the camera correctly.

The key to rotating the camera correctly is knowing the location of the entrance pupil. If you search online you will find many references to the forward nodal point, this is incorrect, the important point is the entrance pupil.

To rotate the camera correctly for a stitched panorama you must rotate it around the entrance pupil.

The entrance pupil is the aperture as viewed through the front of the lens. Often you can see this just by looking into the front of the lens. It is the narrowest hole visible inside the lens.

On a wide angle lenses that hole will appear to sit somewhere between the front of the lens and the camera body. On wider lenses the entrance pupil will be closer to the front of the lens. When my CX-1 is set to it's widest the entrance pupil is roughly 22mm in front of the tripod mount.

On telephoto zoom lenses the entrance pupil will be behind the camera body. When my CX-1 is set to maximum zoom the entrance pupil is over 30mm behind the tripod mount.

However knowing the location of the entrance pupil is not enough, you also need to reliably rotate the camera around that point. For a single row panorama this is relatively easy, I get by with this rig. For multi row panoramas you will need a full panoramic tripod head like the Nodal Ninja. It is worth noting that most of these heads are designed for wide angle lens.

Answer 2

The closer objects are to the foreground, the more you need to rotate about exact pivot point of the lens+body+settings combination, and not behind (as one usually does). Otherwise, you get parallax problems. There is equipment you can buy to help, but I've done it handheld just fine with foreground objects. See here for details:

http://www.panohelp.com/panoramicpivotpoint.html

Manual mode is a must. It is possible to match exposures/brightness after you've taken pictures, but it is a pain. Software can help distortion pretty well.

I recommend a third overlap. Many viewfinders have an AF point or visible bracket around 1/3rd in to help you with this. It's better to overlap than to have something missing.

Answer 3

Choice of focal length is entirely depending on the type of scene to be captured. For a wide panorama perhaps some wide angle lens would be preferable. I have made a few stiched scenes using a 50mm lens (on an APS-C size sensor), which can come out quite nicely as well. Not wide panoramas, but more "normal-looking" images with very large resolution.

I would say that the single most important thing to keep in mind in order to get a consistent set of images to work with (apart from the obvious tripod) is to shoot in manual mode. Take test shots in different directions over your scene to evaluate exposure and when you have decided on exposure, set it in manual mode. This will allow you to get images that have exactly the same exposure, so you don't get ugly edges between the images.

Answer 4

The biggest point is that you want to shoot in manual mode so the brightness stays constant across shots. You don't want to use a polarizer. Though somewhat helpful, a tripod is not really necessary as long as you're careful in your hand-holding. Either with or without, you'll usually get some rather ragged edges that you'll have to crop off. With care, a tripod can reduce that, but it'll happen anyway. You also generally want to avoid particularly wide-angle lenses (especially zoom) lenses -- any lens distortion leads to more ragged edges that need to be cropped.

Answer 5

These are some points I usually follow,

1. Setup your panning plan with Auto or Aperture Priority to meter and focus the primary area of your target picture
2. Then switch to manual mode (don't change settings) for the panning
3. Overlap a third of the neighboring pictures
4. When the objects in the pan are close, avoid too much zoom in your initial setup (however don't go wider than normal -- less than 35mm or some such threshold for your lens and it might start adding distortion that will make joining the images difficult)
5. Additional trick is to take the entire set of pictures 2 or three times with exposure bracketing
   • that way you have the choice of using the best exposure or even going HDR with the pans

I do these things with a Nikkor VR Zoom lens and can get by without a tripod in most cases.