A Better Way To Test Fisheye Lens Resolution
I had tried to measure my Rokinon 8mm fisheye lens resolution a while ago, but I couldn’t get very good answers. The link to that article is here.
The MTFMapper program I use to measure lens resolution is designed to look for little squares in a test chart to make measurements. A fisheye lens distorts the chart squares until they no longer resemble a square (they look like a rhombus shape), and hence the (previous) program versions skipped measuring them.
The MTFMpper author, Frans van den Bergh, has been working on this fisheye measurement problem. He, of course, has come up with a solution (or else I obviously wouldn’t be doing this article, duh). The fix to this problem is to un-distort the photograph of a test chart, so that the little squares in the resolution chart photo look like squares once again. Simple. Unless you had to do it.
I’d like to show some screen shots of the new MTFMapper program, version 0.6.5, that demonstrate how to configure the software to be able to measure fisheye lens resolution.
I always use Raw-format un-sharpened pictures of the test chart(s) that MTFMapper requires for proper measurements. The program uses the “Dcraw” program under the covers, which is constantly updated to understand new “raw” format files.
The un-manipulated 8mm fisheye chart picture (18” away)
Note that the chart shown above is the “vintage” resolution chart. There are newer chart designs available, but the latest MTFMapper version still accepts this older chart design.
Old versions of MTFMapper couldn’t measure the chart sides
As you can see above, the highly-distorted chart was just too much for older MTFMapper versions. Frans worked on adding the image manipulations to (mostly) eliminate the fisheye effect. His program needs to be told by the user that the chart shot needs distortion correction, and this is done in the “Settings” dialog.
Distortion correction in the Settings dialog
Different fisheye lenses use different optical “projection” formulas. The Rokinon 8mm lens uses “stereographic projection”. The options in the dialog include “none” (most lenses), “radial”, “equiangular”, and “stereographic”. If you don’t know the projection formula of your lens, you can just try experimenting with the options. You will need to include the actual focal length of your lens, unless you choose “none”.
The MTFMapper “Annotation” dialog
The picture above shows the measurements of the chart squares in the “Annotation” dialog, and incorporates the distortion correction. Note that it’s not perfect, but the program is able to readily measure all target squares now. The slight remaining barrel distortion doesn’t affect the MTF50 measurement accuracy. The sides of the little squares are now straight enough for MTFMapper to work its magic.
The “Profile” dialog. This is a good lens!
If you’re familiar with typical lens MTF50 measurements, then the above measurements should impress you. Although this Rokinon 8mm isn’t very good wide open, f/8 and beyond show that this lens is capable of amazing resolution.
The “2D” grid MTF50 lp/mm measurements
The resolution “fingerprint” of this lens is quite unique. This is a “DX” lens; the results above are using a Nikon D7100 (3.92 micron pixels).
The “Lens Profile” MTF10 and MTF30 measurements
The new MTFMapper V0.6.5 is a resounding success for tackling the fisheye lens measurement problem. I realize that not that many people own fisheye lenses, so this new feature will probably have a limited audience. If you need it, you know who you are.
I thought I’d mention that I also did an article here that discusses how you can effectively convert your fisheye images into a regular super-wide rectilinear lens (I used Lightroom lens profile corrections). This technique could also get you chart photos that MTFMapper could use, but it’s best to stick with un-sharpened raw pictures when measuring resolution.
If you like this program as much as I do, then please let Frans know!