Distortion test target
What is distortion?
I'm pretty sure, most of you know that zoom lenses show distortion. Usually it start with (more or less) slight barrel distortion at wide angles (short focal lengths) to pincushion distortion in the tele area (long focal lengths). This effect is quite normal with zoom lenses. This does not mean that you lens is defective!
The lenses have been calculated in such a way that some amount of distortion was allowed. It would be possible to create a lens without any distortion, but then the weight and/or price of that lens would be very high. Therefore usually a comprimise is made.
When doing landscape photography you normally will not recognize distortion as long it is not too strong. But as soon you have some straight lines on the photo you'll see it. Lines are not straight any more but bended.
How to remove distortion?
There are several ways to do it. You could for example use a specialized software like PTLens. Some software uses the a,b,c parameters from the Panorama Tools. a,b,c are parameters of a mathematical equotation, which can be used to remove distortion from a photo. (Exact formula: (a*r³+b*r²+c*r)*r r=Radius)
Some software, e.g. Hugin, or Bibble (Pro), uses these parameters to correct distortion in photos before any further work is done on them.
When searching the web you will find several sources describing different methods how to obtain the a,b,c parameters. In my opinion you can easiest do it using Hugin. A nice description can be found here. In that description a taut string is used as a target. But the string is sometimes hardly visible.
Distortion target
To measure the distortion of lenses usually a target like the following one is used:
 |
Such a target is also very usable for obtaining the parameters a,b,c with help of e.g. Hugin. You have clear and straight lines. When talking about Panorama Tools and lens calibration it is oftenly recommended to use a modern building with straight lines (structures). But then you still cannot be sure that everything is really straight. But this is really important for wide angle lenses.
The target is provided in the format DIN A0. I recommend going to a local copyshop and let them print it.
Here is the target in PDF format: Distortion-Target.pdf
How to do it
- Place the target on a flat surface (e.g. a wall). (Especially important for short focal lengths!)
- Mount your camera on a tripod. The lens should roughly be on the same height as the center of the target.
- Choose the distance so that the target fills the frame as far as possible (around 50x to 80x focal length)
- Set you camera to ISO100, Av, F8 (where the lens is most sharp)
- Use a timer and/or liveview to take the picture in highest resolution
- For zooms: Repeat the above steps for several focal lengths
- Load each photo into Hugin and mark several lines. IMPORTANT: Do not mark them als horizontal or vertical lines. Start calculation of only a,b,c parameters. Set projection to Rectilinear.
- Verify the found parameter using your test image and a suitable application (personally I use Bibble Pro). If needed correct the points within Hugin or change the values manually in your application. The values you get from hugin are then the starting point. Hint: When values increase a buckle occurs, when they decrease a dent occurs.
I got the best results when I created three horizontal lines with four controlpoint pairs within Hugin. One line was at the top as far possible, one in the middle, and one at the bottom. I placed the controlpoint pairs from the outside to the center, each one in a separate box. Then I checked the RAW file of the test image within Bibble Pro and most of the time modified optimized the found values by changing the third or fourth position after the decimal point.
The calculated parameters can be used with Hugin or e.g. Bibble or you can distribute them.
BTW: When doing a manual calibration most of the time it is sufficient to find the correct b value. a, and c are only necessary for complex distortions. More informations can be found here.
NOTE: When using the above described method the results are only valid for shorter distances. The problem is that most lenses do not have the same amount of distortion through the whole distance (focus) range. If you are mainly taking pictures of landscapes or object that are far away, you should use instead the above target e.g. a skyscraper with regular window positions (e.g. glas front) as target.
Comments:
