Underwater Bubble Blowin' Users Group

[Tom Hamilton]

Thanks Pawel, lots of new ground for me there to cover :-)

From Wiki;

The modulation transfer function (MTF) represents the Bode plot of an imaging system (such as a microscope or the human eye), and thus depicts the filtering characteristic of the imaging system. The human eye, for instance, acts as a low-pass filter, in that very high-frequency components (sharp edges) cannot be perfectly perceived.

The 'very high-frequency components (sharp edges)' referred to above, the limits of human vision, can that be plotted on your MTF chart above for comparison purposes?

Again, from Wiki;

Factors affecting MTF in typical camera systems

In practice, many factors result in considerable blurring of a reproduced image, such that patterns with spatial frequency just below the Nyquist rate may not even be visible, and the finest patterns that can be seen appear 'washed out' as shades of grey, not black and white. A major factor is usually the impossibility of making the perfect 'brick wall' optical filter (often realized as a 'phase plate' or a lens with specific blurring properties in digital cameras and video camcorders). Such a filter is necessary to reduce aliasing by eliminating spatial frequencies above the Nyquist rate, but in practice, it will have a response that 'rolls off' seriously before the Nyquist frequency is reached.

The 'roll-off' in corner sharpness is clearly evident on your MTF chart, and although the Nikonos displays superior performance, it seems, to me anyways, far from ideal. Where would the 10.25" dome plot fall? Just out of curiosity, what lenses were used behind the flat and dome ports for your test, were they set at their optimum F-stops?

I am beginning to realize that the only solution for the 'perfect' underwater image, may lie in designing a sensor physically constructed to mirror the image field curvature of the intended dome. In other words, a dedicated underwater sensor, just as the Nikonos is a dedicated underwater lens.

The Dome Sensor. You heard it here first folks

[Pawel Achtel]

Sorry to burst your bubble, Tom, but... :)

This has already been attempted by Leif Samuelsson back in 1998. Here is a copy of his research paper: http://home.swipnet.se/marinls/domeuw.html

However, such solution has some fundamental problems other than manufacturing, which Leif pointed out in his research:

The image plane curvature for a given dome port and angle isn't constant. It changes depending on the distance of the subject from the lens.
The objects close to the dome are projected on a surface of a virtual sphere, but objects further away are mapped on a slightly flatter, but also curved surface. Not only those surfaces have different curvature, but they are also not aligned, causing compression of "depth" dimension.

While theoretically possible to overlay, the MTF of human vision would be relatively insignificant compared to that of 4k sensor and normal/wide viewing angles. The acuity of central vision is very high, comparable to about 160 Megapixel array. As we watch images, our eyes constantly scan the area, effectively requireing even higher MTF from corner to corner. So, we are not even close to the limits of human vision. Although 4k may be getting close enough for practical reasons.

A 10" dome is not that different than an 9" dome. CinePort has curvature of 12". You would need a 32" dome to minimise the image plane curvature on a 30mm sensor to similar levels as we had using 2/3" HD sensors with an 8" dome. But, even then, image plane curvature would present problems at large apertures and wide angle of view in 4k.

Not sure why you think that Nikonos is "far from ideal". To me, it seems to be as close to ideal as it gets: it is small and compact, it does not require gears or servos, it can be changed in seconds, it out-resolves the sensor at 5k from corner to corner (not many terrestial lenses can achieve this on land!), it performs exceptionally well wide open, it produces perfectly flat image, it doesnt' cause vignetting, it does not create geometrical distortions, it doesn't flare easily (BTW, speaking of flaring, Have you seen Howard Hall's footage of humpback whales?), it produces high contrast images, it doesn't suffer from chromatic aberrations or astigmatism, the manufacturing was to very strict tollerances and back focal distance is generally spot on, and, best of all, they are inexpensive!

There is also a good selection of different lenses ranging from 12mm 170-degree fisheye to 80mm and anything in between, including all RS lenses that we can use on standard Nikonos mount with an adapter. That's pretty close to ideal to me :) You can even put a dome port and any manual Nikon lens on it for split shots.

But, for me, the biggest advantage is the ability to mout two of those lenses side-by-side or in a wet beam splitter. How do you shoot 3D with two 10" domes?

[Tom Hamilton]

Ok, but the paper you cited has film wrapping around a mechanical guide in order to a achieve the required curvature, and as this can only happen along one axis, laterally that is, the vertical distortions in the frame would still be present. I don't know if constructing a spherical sensor is even possible with present-day technology, but seems to me an interesting solution for the big industry guns to play with.

Perhaps a better way to skin this cat then, a camera that samples varying focus distances near-simultaneously, somewhat along the lines of HDRx, except in focus, which of course, would require a specially constructed lens with very high-speed motors. And a huge amount of on-board memory. And a very complicated software algorithm. Never mind.

The image plane curvature for a given dome port and angle isn't constant. It changes depending on the distance of the subject from the lens.
The objects close to the dome are mapped on a surface of a virtual sphere, but objects further away are mapped on a slightly flatter, but also curved surface.

So how does Nikonos solve for this?

[Pawel Achtel]

So how does Nikonos solve for this?

By ditching the damn "porthole" altogether :)

They were designed from ground-up to produce sharp, undistorted and flat images in the water. Correcting something that was not designed for the intended purpose is generally hard, impractical or often impossible.
What impressed me most was the optical performance and tollerances, which surpass the performance of many modern terrestial lenses on land. I didn't know this before I actually measured MTF.

What does it practically mean? Very crisp, flat, undistorted, high contrast underwater images. My current favourite is this one: http://achtel.com/DeepX/A004_R003_1119WH.001930.jpg . You are welcome to blow it up 1:1 and compare it with any other frame grab captured with any other underwater optics/camera...IMAX included and see what you think.

For me, the cat has already been skinned :)

[Tom Hamilton]

What lenses did you use behind the flat and dome ports for your MTF chart above?

[Pawel Achtel]

Master Prime 14mm at f/5.6

[Tom Hamilton]

Hard to argue with that Pawel, thanks for being a good sport, always knew the Oz boys to be straight shooters. My faith is intact.