Lens distortion analysis

Flame Flame Questions
22

Both of these are good:

The key is to undistort into a larger raster so you have a little slop on the edges that can go back from whence they were generated:

5 Likes
23

@Brooks ^^^ @pbusta3

24

Talk to me about angled lens grids @finnjaeger :pray:

25

You can derive the actual focallength of a lens from angled grids, 3DE can do this.

Basically those are better but way harder to capture, great for things like iPhones and whatnot

3 Likes
26

I like the way Syntheyes does itā€¦ if I understand correctly it calculatesit based on the difference between the 2d trackers and their related 3d pointsā€¦ works quite well for me.

2 Likes
27

It can do distortion that way but cant do focal length/FOV?

I mean it can guess but its not very accurate and depends on the camera move, Ed3 uses the angled lens grids to calculate a focal length thats even more accurate than the marking on the lens (a 50mm might really be 52.3mm and so on).

Or you can use a lidar scan or measured geo to get FOV that way of course when lining up)

there are lots of tricks, and you probably dont need it in 99.99% of cases but to get accurate FOV from weird camera systems this is a nice little tool, its very difficult to shoot full frame angled lens grids as youd need giant charts depending on the lens.

28

One thing that 3DE does, which I havenā€™t seen in PFTrack and Syntheyes (though maybe I just missed it), is the adjust parameter workflow. This is not just a line-based lens distortion analysis, but actually a full simulation of various values to see how they influence the mean error. Not only to fine tune focal length, but other aspects too.

1 Like
29

Agree w the smart folks above.
3de is my solve for this stuff, no disrespect to the syntheyes peeps. Clunky UI and all, 3de has really put time into solving the really tricky stuff w lenses.
Also the matchbox export into flame is nice but I always want stMaps in pocket in case it goes across to other apps.

1 Like