Fully corrective port and wet lenses - Underwater Photogrammetry and camera calibration

Hello,

I'm doing a research about Underwater Photogrammetry, concerning the calibration of cameras for underwater measurements. I'm not a underwater photographer and I'm having trouble to find technical info about underwater ports and wet lenses. This is the reason that I'm posting in this forum, I hope someone with more underwater photography experience could point me to the right direction.

Divers use to say that there are a magnification factor of 1.33 inside water, but the distortion is not that simple. If you want to make accurate measurement underwater you need to calibrate the camera. The air camera model is well know, but underwater, due to the refraction index change, there are some effects that are not very simple to model. The paper Flat Refractive Geometry [1] makes clear that the underwater distortion in flat ports "creates 3D distortions which are not a 2D (radial) mapping of coordinates". The figure below illustrates it.

A scene imaged by a camera looking into an aquarium, empty (top) and full with water (bottom). On the right side, the red stick in the air is
seen as a fence in water. On the left side, a shark is suddenly apparent. Figure 6 from [1].

In the book: Non-Topographic Photogrammetry (1989) [2] I've found that dome ports reduce this refraction problem. But distortions are still present. In the same book it describes a fully corrective port, meaning that the camera see objects as it were in the air. It refers to the Rebikoff Ivanoff lens (1960). Using this kind of port, the air camera model would still be valid, making things much easier.

The best I could find in google is the original article from 1960 "Correcting Lenses for Underwater Use" (I've no access to it) and this link with an image of it:

Rebikoff-Ivanoff correction lens. From [4].

It is odd that there is no recent paper talking about this fully corrective lens or an improved version of it!! At [1], the authors says that the alignment of the lens is very tite, but [2] says that the position of the air lens in relation to the corrective port can vary several milliliters.

Questions:

• Why it is so hard to find information about this fully corrective lens?
• Are the actual wet lenses fully corrective and/or there are correction lens to use in combination with air lenses?
• Anyone know about some recent book or paper about wet lens and/or underwater ports?

References:

[1] Treibitz, T., Schechner, Y., Kunz, C., & Singh, H. (2012). Flat Refractive Geometry. IEEE transactions on pattern analysis and machine intelligence, 34(1), 51–65. doi:10.1109/TPAMI.2011.105

[2] Karara, H. M.; Newton, I. (1989). Non-Topographic Photogrammetry (p. 445). Asprs Pubns; 2 Sub edition.

[3] A. Ivanoff1 and Paul Cherney (1960). Correcting Lenses for Underwater Use.

[4] Website Seafriends. Rebikoff-Ivanoff correction lens.

Questions:

• Why it is so hard to find information about this fully corrective lens?
• Are the actual wet lenses fully corrective and/or there are correction lens to use in combination with air lenses?
• Anyone know about some recent book or paper about wet lens and/or underwater ports?

Answers:

• I guess that they are too difficult/expensive to make, are likely to have chromatic aberration, even if suitable for zoom lenses as approximately afocal, and no-one makes them commercially*...
• ... housed cameras use terrestrial lenses behind the port...
• ... no!**

* there might be an application in scientific or engineering photogrammetry: it could be worth investigating the optical systems used in ROVs.

** commercial manufacturers (Zen, Aquatica, Nauticam, SeaCam etc) design ports, but I don't know if they share their optical data. You could try contacting Jean Bruneau, Viz'art, who is very helpful on Wetpixel.

If the goal is to take accurate measurements based on underwater photographs, distortion is not the only/main problem. You need to know the distance of the object. Some cameras store a distance estimate in the EXIF information based on the lens focus setting (but the Canon 20D and many/all other Canon's don't) but I'm not sure if that is close enough. Using a very shallow DOF lens may help. I've thought about using two GoPro cameras with wide spacing to use the different angles of view to get the distance and, after calibrating on objects of known size, also the size.

If your focus is on understanding the special distortion-free optics this may not be of much use, if you just need to take underwater measurements, tricks like this are perhaps more promising than hoping to have someone build the ultimate port for you.

Bart

Hi Pedro77

You could try Nikonos lenses. They could be way off the mark from what you need. There are literally tons of Nikonos lenses at giveaway prices. Try a few of them and see. From my own experience these lenses are made for water contact and most will not focus topside but UW they are tack sharp and distortion free. The higher end Nikonos RS lenses are even more sophisticated. I have re-chipped a few of them and have added aperture rings and they are probably the best UW corrected lenses there is. They are again designed by Nikon just for underwater photography. I have done some experiments myself. Even made a port that could make the Nikon 14-24mm working without too much corner distortion. Let me know what camera and focal length I may be interested in studying this further.

Cheers

David

A few things to consider:

The 1.33 magnification only applies (approx.) to lenses behind flat glass ports (not dome ports).

A wide-angle lens behind a flat port results in pincushion distortion that gets very severe as the lens gets more and more wide-angle.

A wide-angle lens behind a flat port has significant chromatic aberration (red, blue and green light focuses in different places), making precise measurements dependent on looking at a particular colour of light.

When a lens is placed behind a dome port, the port itself becomes a lens underwater because of the curved air/water interface created by the dome. The dome creates a virtual image of the underwater world, and that virtual image is located at only 3.03 times the optical radius of the dome (from the optical centre of the dome) for objects which are at infinity underwater. To clarify this, to focus on an object which is 50m away underwater the camera will focus between about 0.2m and 0.5m behind most common size dome ports. Focus for closer objects is less than this, so focus extends from the distance from the sensor to the front of the dome for objects touching the dome to a fraction of a metre away for objects at infinity. For a properly matched dome, chromatic aberration shouldn't be a problem. Strong pincushion distortion behind a flat port gives way to less severe barrel distortion behind a dome port. ( http://wetpixel.com/articles/dome-theory ).

Generally, underwater photographers tend to prefer that the entrance pupil of the lens is at the optical centre of the dome, this should minimise distortion.

Because this virtual image behind a dome port is very strongly curved (it's 3.03 times the optical radius of the dome from the optical centre), most rectilinear lenses (which have flat focus planes) will not be able to focus sharply near the corners. Larger domes reduce this problem. Fisheye lenses don't have flat focus planes and so have somewhat sharper corners.

Years ago, Nikon made the UW-Nikkor lenses that may be the best wide-angle solution. The lenses were designed to work underwater (in direct contact with the water) and so don't suffer from the problems caused by trying to adapt lenses for use below water.

Adobe provides tools to calibrate lenses ( http://www.adobe.com/support/downloads/detail.jsp?ftpID=5489 ) , so (at least some) distortions can be corrected in Lightroom or Photoshop. While pincushion and barrel distortions are corrected using this method, it does not appear to be possible to correct for the severe chromatic aberration in Lightroom caused by wide angle lenses behind flat ports. Also while the tools allow the creation of profiles to correct at differing focus distances, if there are large differences, objects extending through the focus plane can't be compensated for completely.

Don't know if that's at all useful.

Hello,

 

I'm doing a research about Underwater Photogrammetry, concerning the calibration of cameras for underwater measurements. I'm not a underwater photographer and I'm having trouble to find technical info about underwater ports and wet lenses.

 

Rebikoff-Ivanoff correction lens.

 

It is odd that there is no recent paper talking about this fully corrective lens or an improved version of it!! At [1], the authors says that the alignment of the lens is very tite, but [2] says that the position of the air lens in relation to the corrective port can vary several milliliters.

 

Questions:

• Why it is so hard to find information about this fully corrective lens?
• Are the actual wet lenses fully corrective and/or there are correction lens to use in combination with air lenses?
• Anyone know about some recent book or paper about wet lens and/or underwater ports?

Concentric dome ports produce a close, curved virtual image which has a complex set of distortions due to the requirement for the wide-angle lens behind the port to operate at a close focus setting which introduces distortions in itself, and the need to use a small aperture in order to maximise depth of field. The dome is also, in effect, a simple lens and this creates aberrations too, and accurate alignment of lens and port may become critical for distortion rendering for high quality photogrammetric use. Concentric Domes can be used for high quality technical work but with a limited field of view.

From memory the problem with the Ivanoff corrector is similar to the problem with a concentric dome port system - both are effectively limited to about an 90 degree field of view after which they become difficult to correct in their image periphery. Ivanoff corector type lenses are rarely found on still camreas these days due to bulk and minimal advantage over conventional concentric ports in pictorial photography.

There have been other designs which have attempted to deal with the problems of wide-angle work and distortion and you might find the (brief) paper "Optics in hostile environments" by Philip J Rogers (35/SPIE Vol. 1780 Lens and Optical Systems Design (1992)) useful as it describes a thick concentric dome used in conjunction with a (complex) corrector lens. It is commented that this is a final high-performance solution but is imperfect in that it requires a large and thick concentric dome. It shows low distortion (<1%) and the optic for which it is shown operates normally (ie focus is unchanged) so will probably perform better than a similar air optic operating at close focus distances. I have no further details and don't know the angle of view but my suspicion is that any design like this is going to be 'lens specific' and will be designed around a particular air optic.

Nikonos lenses are not a solution either and as distortion was not a problem in the pictorial work that they were designed for, they are probably not terribly well corrected - the 35mm lens is a land lens designed for the rangefinder Nikons of the 1950s behind a thin flat glass anyway, the 28mm has a concave front glass and so on. In anycase they can only be used with film to use their full image coverage, and are now old designs. The RS lenses cannot easily be used on digital full frame format as far as I am aware (David - CheungyDiver seems to indicate otherwise) and the digital Nikonos RS cameras produced for the military are as rare as hens teeth, but do apparently exist. Again, whether RS lenses distort is another question - almost certainly the zoom will.

Wet lenses are usually fitted in front of an underwater lens or port and can be fitted or removed underwater. They may provide some degree of solution but I suspect that alignment of them with the camera lens will always be an issue and unless you can determine their optical characteristics and hence how these will interact and operate in conjunction with camera lenses, and hence how accurate alingment needs to be, they may be somewhat hit and miss. Again, I suspect that they will create distortions which may even be severe in some cases but which are not particularly problematic for the applications they are generally put to. Such lenses usually operate on smaller than 35mm format cameras (as do underwater video lenses built by manufacturers like Fathom), so may not actually be ideal for photogrammetric use anyway.

Flat Ports are considered viable at up to approximately a 60 degree filed of view (35mm of 24x36mm format) beyond which they have poor image edges. And they reduce the field of view duue to refraction.

I suspect that a modern aspheric design port could be produced (probably lens specific) but to date this has not been economically viable as a consumer product (the military may have such designs, who knows?). I've looked into ths but the predicted design cost is unlikely to pay back as far as I can see.

Sid Ray's book "Scientific Photography and Applied Imaging" has a chapter on underwater photography which might be worth looking into.

Lastly I would suggest that you need to specify both format and field of view to determine what system is available and to what level of correction it will operate.

"digital Nikonos RS cameras produced for the military"

is this fact or fiction?

I do happen to have converted a few Nikonos RS lenses namely 13mm FE, 20-35 and the 28mm all works with my Nikon D800 with AF and on the Red Epic Nikon mount (manual focus with custom mech. aperture ring).

Could also check this guy out Pawel Actel in Tasmania http://achtel.com/

He has done some extensive test on the Nikonos V 15mm distortions etc UW. Might not be exactly what Pedro 77 is after but not too far off from his quest.

Cheers

"digital Nikonos RS cameras produced for the military"

 

is this fact or fiction?

 

I do happen to have converted a few Nikonos RS lenses namely 13mm FE, 20-35 and the 28mm all works with my Nikon D800 with AF and on the Red Epic Nikon mount (manual focus with custom mech. aperture ring).

Could also check this guy out Pawel Actel in Tasmania http://achtel.com/

He has done some extensive test on the Nikonos V 15mm distortions etc UW. Might not be exactly what Pedro 77 is after but not too far off from his quest.

 

Cheers

I've seen webpix of the digital RS - if not simply photoshopped versions, there may have been a few made/modified with an early Kodak sensor but by now these will be very outdated anyway.

The Nikonos conversion is interesting but....

The 35mm Nikonos lens is a 1950s design and is simply not up to current design standards (not to say it isn't good but it is an old design and thigs have noved on).

The 15mm is much more interesting, although again whether it would meet current photogrammetric requirements (potentially more exacting than red for certain) is something else. It looks to me as though it opimises at much wider apertures than lenses behind domes and this is may just be due to the 15mm being an optic designed for close focus then fitted behind the dome of the lens and perhaps even an optic which works better with a curved image. Again I suspect that the actual design is old and therefore not state-of-the-art but nevertheless if it was correctly desinged it may still outperform dome compromised optics. I've always been sceptical about the 15mm's claimed performance but then it is a stills lens and using it on a red may get around its older design limitations.

The Nikonos conversion is interesting but....

The 35mm Nikonos lens is a 1950s design and is simply not up to current design standards (not to say it isn't good but it is an old design and thigs have noved on).

Are you serious? This is the RS system. The auto-focus may be not but the optics is far better because it was designed only for underwater for full frame. I can say that because I know and I am using them to shoot commercially. There is nothing like it. The 13mm FE, 28mm and the 20-35mm. I am testing the Zeiss 15mm ZE & a big dome and even that can't beat the contrast and sharpness. Anyway it is just my two cents.

Cheers

David

The Nikonos conversion is interesting but....

The 35mm Nikonos lens is a 1950s design and is simply not up to current design standards (not to say it isn't good but it is an old design and thigs have noved on).

 

 

Are you serious? This is the RS system.

I was referring to the original Nikonos lens not an RS lens (these are shown on the site you linked to).

A common misconception is that increasing the 'quality' of the lens behind a large concentric dome will improve overall image 'quality'. It might improve central image quality but not in the edges nor corners as the problem here is the curved virtual image being too close which is an optical problem caused by the dome and a better lens won't deal with it. Nikon, Canon, Zeiss; all are limited when the angle of view becomes to wide by the dome itself. There is no workaround.

And I still suspect that the RS lenses distort which might make them problematic for photogrammetric use - they are after all SLR lenses and the wides will have probably some degree of distortion as a result.

Hello everyone, sorry for the delay, I was travelling. I’m very glad for having so many answers, some of them very detailed. Thank you all!

Let me try to compile all the info provided by you:

Fully corrective lenses

• Too expensive
• No one makes it commercially (really?)
• Have Chromatic aberration (not a problem to me, see bellow why)

Try to contact commercial manufacturers: Zen, Aquatica, Nauticam, SeaCam

Try to contact: Jean Bruneau, Viz'art, who is very helpful on Wetpixel.

Ok I will.

Try Nikonos RS and UW-Nikkor lenses

Are theses lenses fully corrective? As in air?

Some clarifying:

I still have not defined the camera, angle of view or distances. Now I’m mainly concerned on the state of the art in underwater lenses and photogrammetry.

Manta Ray,

Why do you say that I need to know the distance to the object? To correct the distortion based on it as [1]? In photogrammetry, it is common to use some known objects to have an absolute value reference. If I was only interested in the object distance, a good way to measure it is using a laser triangulation system. I’ve never thought of using the camera zoom and a short DOF to measure the object distance, it is a nice idea, but I’m not sure how accurate it can be, probably only a good estimation.

Image distortion (“2D distortion” is not a problem)

To clarify, image distortions (pincushion, etc) are not the main problem when they are merely 2D distortion, like a wide angles lenses in air. This kind of distortion can be calibrated/mapped.

The problem is the “3D distortion” caused by the refraction on the water, in this case the camera cannot be described by the SVP (single view point) model. The distortion and the focal length depends on the distance of the object, as you can see at figures below:

http://i.imgur.com/lYbdcXn.png

Image From [1]: “Fig. 4. Looking through a flat interface into a medium yields a non-SVP system despite the use of a perspective camera.”

http://i.imgur.com/6pgEHOM.png

Image From [1]: “Fig. 5. The distortion caused by refraction is nontransversal and depends on the object distance. When embedded in water, both the square and the round objects are projected into the same coordinate. However, in air (the undistorted projection), the objects are projected into different coordinates. Thus, calibrating radial distortion according to one of the objects results in an error when rectifying the other. This happens because the distortion is nontransversal and depends on the object distance in addition to the radial coordinate.”

More about the Fully Corrective Port

The image below from [2] shows the optical path for a fully corrective port in combination with an n-air lens. As you can see, the camera “see” the object as there were no water. The corrective port should be custom made to a specific lens, it is not a problem if the lens have 2D distortion, because it is not due to the water, so it does not depend on the object distance, meaning that it can correct through calibration (as some softwares you pointed out does).

Based on yours replies, this corrective ports are hard to find. It is odd that there is no new info about it.

http://i.imgur.com/YYqj7qGl.jpg

Removing Chromatic Aberrations

I can get rid of the chromatic aberrations using a wavelength filter and using a light source of the same wavelength. Sharp images in the price of no colors, but I do not need them to measure shape.

Hi Pedro

I am making some guesses here. How about using a 50mm prime lens? This is closest to what a human eye would see. Lots of fast and sharp 50mm primes around and cheap too. 50mm is easier to find a dome port. Flat port you will get 25% magnification from refraction from he port.

regards

David

The Nikonos lenses are of the fully corrective port design. If you grab a Sony NEX camera and a Nauticam housing you can get an adapter from Nauticam to fit the old UW-Nikkor lenses. They are manual focus lenses.

There is a 15mm f/2.8 (23mm equivalent), a 20mm f/2.8 (30mm equivalent) and a 28mm f/3.5 (43mm equivalent). There are also 35mm and 80mm lenses, but these are of a different design that use a flat port and therefore suffer from the problems you mention.

There are also the newer autofocus R-UW-Nikkor lenses. These have been adapted to work with Nikon cameras in some underwater housings. There's a 28mm f/2.8 and a 13mm fisheye lens. (There is also a 20mm - 35mm lens but this will not work with any underwater digital camera housing.)

see: http://wetpixel.com/articles/nikonos-rs-lens-project-finalized

and: http://abelic.net/wp/

CheungyDiver,

thanks, but I'm not looking to buy any lenses now. I'm looking for technical data. I want to understand more about underwater ports, lenses and fully corrective lenses (FCL). It is hard to find info about the FCL. tdpriest said that they are too expensive and that nobody makes it commercially, but if [2] is correct, they have a lot of qualities. It is odd that there is no recent article/book/website talking about it.

coroander,

Are you saying that the Nikonos lenses are fully corrective lenses? Thanks for the links.

Pedro, check this link, it should provide useful information http://www.cinemarine.net/cinemarine/fichiers/File/optiqueus.pdf, while the topics of photogrammetry is not approached in this document, there is quite a bit of info on corrective optic for underwater imaging.

A system we have provided for underwater photogrammetry recently for this purposes, will uses 3 full frame Nikon Cameras simultanneously triggered and equipped with 20mm Focal lenght prime lens behind large dome, lighting being provided from a constant source and multiple laser permanently fixed to the rig, will give the proper necessary alignment and distance data.

You can also PM me if you have other questions.

It also available in French, which is a more complete document, especially regarding the Correcteur Asphérique THALACETOR (aspherical corrector), I don't know if anyone in your team is fluent in French, I doubt this would fare well in Google translate :-)

It also available in French, which is a more complete document, especially regarding the Correcteur Asphérique THALACETOR (aspherical corrector), I don't know if anyone in your team is fluent in French, I doubt this would fare well in Google translate :-)

Viz'art, thank you very much for the pdf! Some equations are not completely explained nor referenced, but that is ok. The most interesting part starts at "14 Correction systems". Unfortunately there is no much detail nor bibliographic references or commercial equipment samples.

There is no much info returned by googling Thalacetor, what I've found is:

http://www.cinemarine.net/cinemarine/rubriqueus.php?id=10&page=77

It seams that this guys sells it:

http://extrem-vision.com

Video, I think it features François Laurent,

http://www.extrem-vision.com/objectif-plongee-thalacetor/

Do you have more info about the THALACETOR or any other fully corrective port?

I wounder why the commercial wet lenses aren't already aspheric fully corrective ?

Pedro77. There is nothing new, in that aspheric ports have been around for a long time. Cousteau apparently used them back in the 1950s - one off's ground specifically for the 16mm camera lens he was using. I think that Fathom lenses are fully corrected ports but no doubt users will comment.

What you need to consider is that its 'relatively easy' to build fully corrective ports/lenses if the format is relatively small and/or the resolution requirement's are not too taxing. But trying to build one for a larger format (35mm for example, and high resolution (20MPixels+) and the physical size, cost and optical problems will escalate. Photogrammetry has been the province of medium and large format cameras above water and has required low distortion lenses (I remember using a SWC with the 38mm Biogon for such work as a student - its characteristics were well documented and its distortions well known). If you limit your researches to relatively small formats then you may find more information but shifting to 'photogrammetric formats' will probably yield a great deal less.

Also be aware that whilst you may find some quite detailed optical formulae on the web about dome ports and their optics, most that I have found fail to take the fact that lenses are not working at infinity into account - which will result in more complicated fomulae, especially when a wide field of view is required. Also you need to appreciate that the traditional way of using a low distortion near symmetric design is no longer viable due to problems with digital sensors and acute incidence angles (look up Leca M red edge problems which will explain - I have an old 1960s 21mm Super-Angulon which suffers if used on 35mm full-frame cameras.

Its interesting that Jean has commented about a system equipped with 20mm lenses because this has been the focal length considered as providing 'acceptable' corners on full frame 35mm format for a very long time. You may not find lots of data on this, but the rig Jean describes sounds like one derived from empirical results to provide optimal results from an acceptably wide angle lensed system.

Paul Kay,

thanks for the info, but I need some references (lens manufacturer, paper, magazine, book) to keep my research. I know that the corrective lens are old, this is why it is odd that I can not find new info about them.

I understand that for larger sensors and/or more resolution the optics gets more complex. I want to know what is available on the market. Where are the new 2010's fully corrective lens versions of the 1950's??

The Fathom lenses website says its lenses are "low distortion". So, not fully corrective?

Also you need to appreciate that the traditional way of using a low distortion near symmetric design is no longer viable due to problems with digital sensors and acute incidence angle

Sorry I didn't understand this part, what you mean?

Similar question at: http://wetpixel.com/forums/index.php?showtopic=4442