Stitching and cropping

War memorial, Serbia, March 2007

I read the recent tutorial/essay on framing and cropping from George Barr over at D.O.P. this week and figured he might have missed something.

George does a lot of multi-image stitches, as I do. It's an effective way of getting higher resolution or wider angles with digital. There are a couple of problems, however, with framing related to stitching in this way.

Firstly, there is the problem of visualising the scene, as mentioned in the article. If you are doing tight zooms on a small area then that may not be a problem. Effectively you are replacing a wide angle lens with several narrow angle images. Handheld framing devices would work quite well (given a caveat about point 2, below). For large panoramas, though, this is going to have a limit: if you are photographing beyond human visual field then a framing device isn't much use.

Secondly there is the problem of wastage from the stitching process. Good stitching takes into account lens distortion and perspective distortion, warping the images to produce the final view. This inevitably causes some wastage around the edges of the image due to curved components: cropping is necessary.

To over come thee issues I follow a different process of framing and composition when I'm looking to stitch several images together. This method is stems from a few failures and quite a lot of trial and error. There are 2 steps, in line with the 2 problems.

Step 1 is to determine the elements I want in the final framing. It is about pre-defining a set of edge points. This is much as one would do with a framing device but can be done for very wide angles of view.

Step 2 is to ensure that I add some extra to each frame I shoot to cover the stitching distortions. That often means surveying the whole scene to be photographed to ensure that I define an outer boundary around the whole scene. Typically 10-15% extra needs to be added top and bottom for a left-right pan (and vice versa). I don't take into account view-finder margin in this either, that gives a little more fat. This means shooting slightly wider for each shot than ultimately desired. If this means going wider than you would like for the resolution sought, then multiple rows are in order.

In summary: determine the key elements to be included in the final image, shoot wider to allow for stitching loss, shoot multiple rows if necessary.

The accompanying shot is an example: made from 6 images, handheld, the full size picture is 4200x7800 pixels.


In a side note, George Barr has posted a reply to a comment I made about knowing when to keep working on a picture. A very good post that helps consolidate my thinking.