Deck chair left and right: White stripes have a green fringe on the inside and purple on the outside due to chromatic aberration
Another problem that almost all lenses produce is chromatic aberration. Again van Walree has one of the better explanations. Whereas radial (barrel) distortion will interfere with panorama production because the images diverge very quickly from one another – thus making it necessary to have a sharp transition from one photograph to the next or risk having highly divergent control points, chromatic aberration will interfere with the identity of control points in highly contrasting areas. Consider the transition from white to black to white on the left hand side of an image: The black line in the middle is fringed green on the lefthand side and purple on the right. On the other side of the image a black line has exactly opposite fringing. If the line is thin enough it may appear as a green-purple line on the left and purple-green on the right, making it difficult to establish it as a control point.
As I say this is not only a question of how much money you’re willing to put up front to solve this problem; it is more a matter of how to deal with it when it arises.
The common graphics programs all have functions that correct chromatic aberration and I have explained how to do this in a tutorial. All that is needed in most cases is to shrink the blue and the red channels by a small number of pixels so that they fit onto the green channel and the fringing is gone. Now if we could find out those values for a particular camera, then all our problems would be solved.
Turns out that this is not done just by determining two values. Chromatic aberration is generally regarded as being inversely proportional to the f-stop value, i.e. the higher the f-stop (8, 11, 16…), the less chromatic aberration there should be. As far as I can see, however, this is an untested assumption. Furthermore, chromatic aberration can also be expected to depend on the zoom setting, but in an unpredictable way.
Now we have a burgeoning problem. In the Canon SX series there are 23 zoom levels and 13 apertures, making a total of 299 combinations; and apart from an unconfirmed rule of thumb, there is no way of estimating even the ballpark area where the values might be found. Of course, we could try individually correcting each image, but that defeats the point of the exercise: Digital images are cheap and heaps of data are waiting to be collected and processed. Only, if we had to then do this by hand, we’d just end up creating work for ourselves.
Leaving aside for a minute that problems might be confounded by chromatic aberration depending also on the distance the camera is focussed at, is there any reasonable way of even collecting and analysing the data?
Turns out that there is, but, again, it depends on an unexpected twist to the story.
