How not to copy a picture

[Dick Roadnight]

Clauss pano picture copying at Assisi
Pano heads have their uses, and I am thinking about getting a Clauss, but in this instance they seem to have got everything wrong:

They use a DSLR with focal plane shutter, mirror and Anti-Aliasing filter...
... and not a High-res AA-free, mirror-free professional MFD camera with an electronic shutter.

They even use a low-res Nikon D2X for one picture, but, if you combine enough images, you can get as much res as you want.

They use a lens designed for sport or wildlife work, not copying.

For recti-linear work they use a polar/pano system...
creating perspective distortion that will have to be rectified in software, resulting in further loss of resolution.

A Cartesian rectilinear system, keeping the sensor parallel to the picture ¿zig-align? would prevent this distortion... a normal "move and stitch" system easily achieved with a Cartesian robot (rather than a polar/pano/rotating one).

Moving the lights with the camera might have some benefit... but generally, when copying paintings, you need to keep the light well away from the camera (especially when copying oil paintings)... but, you might get away with it for matt art work.

If you wanted to study the brushwork of Leonardo, you would need oblique light, so, while you were doing the job, you should take two sets of pictures, one showing the brushwork, and one for the picture.

Their set-up picture shows perspective distortion, visible in the back wall, which makes the arches look asymmetrical.

The Italians (and Germans) are generally fully up to speed with technology as well as art.

[Sheldon N]

I would think that using a smaller format camera like the D2x with its high pixel density combined with a long focal length lens like the 600mm is the perfect choice for this type of application. It appears that their goal was to extract as much resolution out of the original as possible through stitching, and I'd say they succeeded. If you click through to their final result and zoom in, it's astonishing how much detail there is... up close with a magnifying glass and your nose against the fresco kind of detail. I don't think that you could accomplish the same result with a MFDB as easily because of the lack of long focal length lenses and lower pixel density.

Not sure about the perspective distortion issue, but the image of the Last Supper looked fine to me.

[Dick Roadnight]

I don't think that you could accomplish the same result with a MFDB as easily because of the lack of long focal length lenses and lower pixel density.

With a MFDB you could use a Apo Digitar macro at 1:1 or more and get any amount of detail you wanted.

[Bart_van_der_Wolf]

Clauss pano picture copying at Assisi
Pano heads have their uses, and I am thinking about getting a Clauss, but in this instance they seem to have got everything wrong:

They use a DSLR with focal plane shutter, mirror and Anti-Aliasing filter...
... and not a High-res AA-free, mirror-free professional MFD camera with an electronic shutter.

Why would one prefer aliasing artifacts that cannot be removed afterwards over antialiasing measures that can be removed? With a stable positioning system, I also don't see what in their choice of equipment would hurt image quality.

They even use a low-res Nikon D2X for one picture, but, if you combine enough images, you can get as much res as you want.

They use a lens designed for sport or wildlife work, not copying.

Exactly, they used a high magnification (small angular view) device from a distance, with a very good MTF over it's useful FOV, and little color cast across the sensor plane. Using a larger sensor array would have made it possible to use fewer tiles.

For recti-linear work they use a polar/pano system...
creating perspective distortion that will have to be rectified in software, resulting in further loss of resolution.

The result apparently has a 20 pixel/mm (508 PPI) resolution, plenty of leeway for small (less than 1-2 pixel) deteriorations when printed at true size. The deterioration from resampling such a limited angle of view to rectilinear, may well be better than the blur that a single lens' field curvature can produce. Each tile was presumably focused optimally across the field of view. Besides, stitching will usually mostly use the center of the image circle for each tile (the edges/corners are lost in blending).

A Cartesian rectilinear system, keeping the sensor parallel to the picture ¿zig-align? would prevent this distortion... a normal "move and stitch" system easily achieved with a Cartesian robot (rather than a polar/pano/rotating one).

Sure, but perhaps there were restrictions for moving side to side and up and down from the shooting position. By shooting from a single position, they could simplify the camera setup (faster setup time, and faster sequential shooting).

Moving the lights with the camera might have some benefit... but generally, when copying paintings, you need to keep the light well away from the camera (especially when copying oil paintings)... but, you might get away with it for matt art work.

Agreed, that part puzzled me as well.

Cheers,
Bart

[Dick Roadnight]

... stitching will usually mostly use the center of the image circle for each tile (the edges/corners are lost in blending).
Bart

Move-and-stitch uses the center of the image circle, and if you use a 53mm * 42mm 60Mpx sensor on an Apo-Digitar with a 100 or 150mm image circle, you only use the center of the image circle anyway.

[JoeKitchen]

Most museums that I have heard of use multi shot backs as well, taking away the need for the camera (or computer) to interpolate, producing much sharper and accurate edges.

[Dick Roadnight]

Most museums that I have heard of use multi shot backs as well, taking away the need for the camera (or computer) to interpolate, producing much sharper and accurate edges.

...and you don't get any more pixel count (file size is no bigger) but you get more res than with a single shot DB, and much more res (per pixel) than you would with a camera with an Anti-Aliasing filter.

[fredjeang]

...and you don't get any more pixel count (file size is no bigger) but you get more res than with a single shot DB, and much more res (per pixel) than you would with a camera with an Anti-Aliasing filter.

He he...that is why the Leica M9 has a better definition than the CaNikons for a same format. Specially looking in little details (trees, grass, hair etc...).

[feppe]

He he...that is why the Leica M9 has a better definition than the CaNikons for a same format. Specially looking in little details (trees, grass, hair etc...).

One man's perceived sharpness is another man's aliasing artifacts...

[Sheldon N]

With a MFDB you could use a Apo Digitar macro at 1:1 or more and get any amount of detail you wanted.

Not exactly a practical solution. They are shooting wall mural segments that are very large. To shoot and stitch  a 1:1 macro of a 15 x29 foot mural (the size of the Last Supper) would require some extremely complicated mechanisms and framework to move the camera around and keep it close to the wall. Much simpler to back the camara way off, use a long lens and shoot spherical panos, then correct for any distortion in post processing.

Agree that a MFDB is a great choice for shooting smaller artwork or other items where you are shooting for ultimate single frame image quality. However, if you look at it in the context of their approach (seeking maximum detail resolution for very large subjects, and doing it by stitching) then their choice of a long lens with a high pixel density sensor makes the most sense. They shot it with a 600mm lens which is longer than lenses commonly available for MFDB, so their end resolution is going to be much higher - regardless of AA filters or correction for distortion.

Have you looked at their final result from the Last Supper project? Click around and zoom in, the level of detail is astonishing. I don't see how you can say that their implementation or end result is flawed. (I do agree that the lighting could be better though.)

http://www.haltadefinizione.com/magnifier.jsp?idopera=1

[Dick Roadnight]

With a MFDB you could use a Apo Digitar macro at 1:1 or more and get any amount of detail you wanted.

Not exactly a practical solution. They are shooting wall mural segments that are very large. To shoot and stitch  a 1:1 macro of a 15 x29 foot mural (the size of the Last Supper) would require some extremely complicated mechanisms and framework to move the camera around and keep it close to the wall. Much simpler to back the camara way off, use a long lens and shoot spherical panos, then correct for any distortion in post processing.

For 20 pixels/mm, or a great deal less for the same res using a MFD Cartesian system, you get a square foot or two per capture.

An auto pano head is more complicated, but known technology for photographers.

Cartesian scanning systems are also known technology.

Agree that a MFDB is a great choice for shooting smaller artwork or other items where you are shooting for ultimate single frame image quality. However, if you look at it in the context of their approach (seeking maximum detail resolution for very large subjects, and doing it by stitching) then their choice of a long lens with a high pixel density sensor makes the most sense. They shot it with a 600mm lens which is longer than lenses commonly available for MFDB, so their end resolution is going to be much higher - regardless of AA filters or correction for distortion.

For large cylindrical or spherical subject with a pano head e.g. landscapes, a long lens gives you more pixels per square degree and a pano system gives adequate results. For flat artwork you do not need long lenses for high res.

They make a laser-guided system for getting dead straight lines marking out football pitches, and you could use that technology for pictures (or buildings) too large for a single Cartesian robotic frame, but a 30 foot frame or rail would not be a problem.

[rueyloon]

If you build (or use) a fancy, highly technical and technologically advanced (looking) rig, you will get the job. Most likely the image they produce is good enough for any use by the client. Only photographers like us here would be splitting hairs and discussing why another lens or camera will do a better job.

[Dick Roadnight]

The move-and-stitch technique is widely used for copying flat work, where there is not sufficient image circle for shift-and-stitch, and it is, of course, the standard technique for aerial photography.

For aerial photography you might want to be able to make adjustments to make the images line up, and compensate for non-perpendicularity.

Pre-digital you could correct for distortion without losing res by using base-board tilts when enlarging.

With copying you would like to use an accurate system that needed no correction, saving post-time, distortion and loss of res... this could be achieved by using a very precise mechanical system or by using a laser system to correct for any misalignment pre-exposure.

When photographing an 800 year old wall in an earthquake zone, you would not assume that the wall was perfectly co-planar, and you could use a slightly flexible rail, and use a simple ¿Leica Disto? laser measure to  align the rail with the picture, but the laser fine tuning system would be easier and cheaper and better I think.

There are six degrees of freedom to consider, and it gets complicated when you try to work out what you are trying to do - if there is a misalignment in the middle of the picture do you record it like it is or e-restore it to look like it did when it was painted? This decision could be made in post, and (whatever the client wanted) you would try keep your options open so you could do both.

[Dick Roadnight]

If you build (or use) a fancy, highly technical and technologically advanced (looking) rig, you will get the job. Most likely the image they produce is good enough for any use by the client. Only photographers like us here would be splitting hairs and discussing why another lens or camera will do a better job.

...so what you need to be a successful photographer is marketing BS, not technical know-how!

[Ben Rubinstein]

They did the job successfully while you talk about how X equipment could have done it better. Personally I believe they have more credence than gearhead talk...

[feppe]

...so what you need to be a successful photographer is marketing BS, not technical know-how!

I think the point was that there are "good enough" solutions which might work faster, safer and/or cheaper than the optically optimal alternative. That good enough -concept is good enough to build bridges, so I'm sure it's good enough for photography.