Optical distortion correction without a dome port

General Discussion
1

Hi,

I’ve noticed lot of subsea cameras with water corrected optics use Ivanoff style correctors, which give a flat port on the outside. An ivanoff adapter will correct the distortion from the water from behind the flatport (and it’s much easier to protect a flatport than a domeport).

Optical distortion correction
Optical distortion correction is really only important for photogrammetry.
Software distortion correction isn’t ideal, as the k distortion values change with distance from the subject when not optically corrected.
The only ways to correct optical distortion seem to be with a dome port or an Ivanoff-Rebikoff style corrector.

On domes,
The acrylic domes are good, but people say they add aberration, and if you want to add a decent camera and lens, you might need a 6 inch dome to get the virtual image far enough from the lens to allow good depth of focus (the dome creates a virtual image of the subject).
Glass domes offer the sharpest corrected image (over both ivanoff correctors and acrylic domes), but a dome that size on the ROV makes me nervous that it would be easily scratched (or even smashed).

Ivanoff corrector
I’m very keen to create a camera that would use an Ivanoff lens to correct the optics, and a box camera inside a blue robitics 4 inch enclosure. But it’s difficult to find somebody to create the lens, especially for a low budget one-off build.

If anybody knows of somewhere these can be ordered or bought from, or any kind of commercially available product, that can be used instead, I’d be keen to know about it.

Also, if anybody knows of any alternatives to optically correcting distortion caused by water, I’d be keen to know more about it.

Happy to share my experience here, to help grow the community.

Kieron

5 Likes
2

Hei, a late partial answer

I’m from Abyssar, a startup working on water-corrected lenses for small sensors.

The Ivanoff corrector, yes it can improve the image quality and distortion, but has other problems.
Take as ballpark max 60 degrees of FOV that Ivanoff can handle well.
For narrow angle (say, 30 deg) you can make yourself a decent corrector with one glass meniscus taken from air-spaced telescope objective (say 80mm diam) and one flat-convex lens. If you’re in the US there is SurplusShed!

Better option, you can install a “Wide wet lens” like those for camera cases, it does the same job; if diameter is ok.

A third option is to find a wide angle objective with slight fisheye distortion, and put it back a flat window; with some luck the two distortions will cancel.

For photogrammetry, my understanding is that distortion can be corrected digitally with decent results. Actually I personally advocate full-distortion fisheyes because it makes more sense optically, you measure angles (the photogrammetrists may disagree).

There are always some troubles when measuring very near objects, as the directions get a bit wacked.

In the end, everybody wish for an objective super-wide-angle-and-sharp-to-the-corners-and-with-zero-distortion-and-f/1.4-and-macro… how much will it cost…

2 Likes
3

Hi Giorgio,
Thanks for your response.

A domeport is our baseline solution, but it leaves a lot of glass to protect.
A 6.5 inch domeport seems to be required to give a sharp image 3 radiuses from the dome centre, and a 6.5 inch domeport requires a very large skid in order to protect the glass adequately - especially with a wide angle lens.

I have heard of people using fisheye lenses can you explain why you believe fisheye lenses would be better?
I haven’t tried using a full distortion fisheye for photogrammetry, except for a go-pro which didn’t produce great results.
I’d be keen to understand your suggestion better, have you any examples of the kind of lens you suggest?

Photogrammetry does digital distortion correction, but underwater the k values (distortion coefficients) vary with distance from the subject. So calculating a single set of k values is impossible and leads to broken and inaccurate models. The only solution is get too much data and set variable k values, which isn’t ideal.

The only camera’s I’ve been able to get reliable photogrammetry models produced from are dome port cameras and cameras with an Ivanoff corrector (these cameras are often very expensive).
I prefer the ivanoff corrector because of the size and ease of protection.

Any more info you can share would be greatly appreciated.

Kieron

1 Like
4

Hi Kieron

that’s a lot of stuff, and you’re not the only one with those struggles. See for example the works of Menna & Nocerino, they are trying to sort out the mathematics core for distortion correction etc.
https://www.mdpi.com/1424-8220/16/1/48

Fisheye in my opinion is good because it sends angles to pixels; while non-distortion wide angle sends a flat target object length to pixels, which is an hefty unnatural job, with its angle accuracy actually increasing for wider angles, with the target getting farther.

The distortion coefficient varying with distance, that’s caused by the “Entrance pupil movement”, meaning the “point of view” is not a point, but changes with the direction.
This issue is common in wide angle lenses and gets much worse with underwater ports and domes because they are large, so the shift in “point of view” position gets larger (several mm), and whacks the apparent direction for near objects.
In principle this issue can be measured and corrected, but in practice there will be daunting mathematics and adjustments as you wrote.

There are ways to solve this problem at the root, with a "fixed entrance pupil “point of view”. I’ve promised such objective to myself about a year ago, but things move slowly. It is a “Probe” lens, like the Laowa 24mm probe or the Nauticam EMWL system.
The “entrance pupil movement” is much reduced because with this Probe configuration it becomes possible to make a very small diameter front port.

As said in previous post, I’m working now actively on design and manufacture of miniature Ivanoff ports and native underwater objectives. Most troubles with ports, domes and Ivanoff arise because we put in the water objectives designed to work in air.
https://www.abyssar.com/

As of now I work only for small sensors (1/2.8"). If you use small sensor or even better the Raspberry camera, in a couple of months I’ll have some ready. 15mm port diameter :wink:
For large sensors like FF 35mm, the best option may be to get an old Nikonos objective and adapt it.
Or wait a couple of years and we’ll remake the Nikonos optics if there is enough interest.

2 Likes
5

we can continue the discussion here for the sake of sharing those common issues

A simple homemade Ivanoff corrector (as the original model by Ivanoff - Rebikoff…) can be ok for narrow FOV like 30 deg; but I’m not sure how it will fare with exact distortion and the above issue of “entrance pupil movement”.

One more “makeshift trick”:
You can use a small dome - say 2 inches - over a smartphone objective; for example the one of Raspberry camera 3 (little metal rectangle, Autofocus)
It works decently because the smartphone objective is built as “pinhole” or “pancake”, without the large front lenses typical of camera wide angle objectives.
It needs to be dead-center in the dome, calibration will be accurate only for one focus position, and has some substantial field curvature (the focus is different for center and corners).

Are coming out some new, very small “pancake” wide-angle objective, like Canon RF 16mm 2.8 and 28mm 2.8, that may tolerate working with a small dome; like the smartphone objective above; i haven’t tried them yet.

2 Likes
6

I just want to address the issue of entrance pupil movement.

In subsea environments, there is not such a need to pull focus in and out.
I’ve always found it best to fix the focus at a set distance and not change it.

Sounds counterintuitive, but it eliminates focus hunting. And by using a smaller aperture it’s possible to get everything in focus with the right camera and lens, so you’re better off not allowing the focus to change.
So I’m not worried about entrance pupil movement negatively affecting focus.

With an Ivanoff port, the field of view obtainable is supposedly in the order of 80deg for an aperture f/2, with a depth of focus double than that of a flat port.

@Patta - I’d like to understand why you say an Ivanoff corrector is only ok at smaller FOVs, I’ve seen wide angle Ivanoff correctors before as part of homemade projects and on a number of commercial ROV camera. Imenco Orca, SubC Rayfin, and one other market leader (I’d rather not name). I’ve worked with all 3 cameras and know they use Ivanoff style correctors, also they’re much easier to protect than a glass dome.

If anybody has experience making ivanoff correctors, or knows why they’re not popular outside of commercial ROV cameras, I’d really like to understand.

Also, if anybody knows what is required to get a lens made, and if there are companies that can make lenses on a small scale? I’d really like to make one of these as an endcap of a blue robotics housing.

FYI I found this link to the original Ivanoff paper describing the lens dimensions he used. I’m sure these have been further refined by Rebikoff and others but I can’t find any other papers with lens dimensions.

Also check out this guy, he put an ivanoff corrector on a 20mm lens.
Note the flat port, but undistorted images of the pool.

1 Like