Gigapan system and shooting flat art

[Roscolo]

I'm doing larger and larger giclee' prints, so I need more resolution. Thinking of getting a 4x5 scanning back, but curious if anyone has used the Gigapan system to shoot flat art for to create high res files to create reproductions of paintings?

[elf]

I'm doing larger and larger giclee' prints, so I need more resolution. Thinking of getting a 4x5 scanning back, but curious if anyone has used the Gigapan system to shoot flat art for to create high res files to create reproductions of paintings?

Taking the pictures can easily be done with the Gigapan sysytem, or for that matter, any spherical pano head.  The problem is in the stitching software.  The software creates the final image as though it were projected on a spherical screen. You can have the software, PTAssembler in particular, use other projections.

Instead of the Gigapan or other spherical pano head, I'd recommend building a track system that moves either the artwork or the camera parallel to each other.  

[jeremydillon]

I've had a crack at this ... thought it would be a great solution but....

1.  It is amazing how much depth of field is a problem.  Shooting at f16 would not get me sharp results shooting a 2m canvas from 5m! (without refocussing each image).  Refocussing each image is slow and tedious and introduces its own problems in stiching as the image magnification changes as you focus. Even if i did refocus for each shot, the frames towards the edge of the painting would suffer depth of field issues.

2. If you are cross polarizing be aware that you will need to reset your camera polarizer as the camera moves. (This is not a problem if the lights are mounted to the camera platform and move with the camera) I ended up using a large sheet of polarizer on a c stand in front of the camera, not attached to the lens.

3. The stitching programs i tried were rarely able to automatically stitch the paintings together, forcing tedious control point marking for each image.

I now on the hunt for a reasonably priced scanning back


PS The projection method is not the problem. All the stitching software I tried offered a recilinear projection.

[Craig Murphy]

I had a Sinar 54H system quote for $15,000 at one point that included a Mamiya AFD, 80mm,  the back, computer system.  All used of course.  Museums are using the Sinar 54H to shoot artwork for reproduction.  The 54H is a one shot or multiple shot camera.
Jeremy.  Why would you have to refocus if the camera was traveling parallel to the artwork?  Were you referring to the Gigapan method?

[JonathanBenoit]

I had a Sinar 54H system quote for $15,000 at one point that included a Mamiya AFD, 80mm,  the back, computer system.  All used of course.  Museums are using the Sinar 54H to shoot artwork for reproduction.  The 54H is a one shot or multiple shot camera.
Jeremy.  Why would you have to refocus if the camera was traveling parallel to the artwork?  Were you referring to the Gigapan method?

The Hasselblad multishot, coupled with the Phocus software there isn't a better option available. Gigapan is not for artwork reproduction. The camera has to be parallel to the material.

Ideally, you would have a motorized slide with a multishot back on a view camera with a 72mm Schneider with automatic shutter. You would also need some pretty reliable strobes with fast recycle time.

You are looking at $40,000 at least. This is if you want to do it correctly.

4x5 scanning backs are becoming obsolete, and for good reason. Better light has ceased their production and will only service them going forward.

[Roscolo]

Apparently it can be used for this. I found a few examples on the Gigapan site. I don't know how much time was invested in stitching or correction. I don't want to have to do an insane amount of stitching. The spherical effect seems to be easily correctable in Photoshop. In analyzing these examples, focus or depth of field does not appear to be a problem. These look pretty spot on to me. jeremydillon: What camera / lens combo did you use with the gigapan system when you tried this?

Polarizers won't be an issure for me. I've photographed artwork for years for artists and museums and almost never used polarizers. In my experience, polarizers always seem to introduce more problems than they solve.

Here's three examples on the gigapan site:

http://www.gigapan.org/gigapans/19770/

http://www.gigapan.org/gigapans/41714/

http://www.gigapan.org/gigapans/19762/

The lighting on the second example is atrocious, but it is a good example because of the 3 examples, it is the only one giving the size of the original painting: 116cm x 176cm, or about 4 ft. x 6 ft. That would seem to mean a lot of angle, and I see no ill effects from using the gigapan system. Looks square and I see no traces of stitching or misalignment. Focus looks sharp even on the edges and corners. If one can achieve such results reliably and consistently without too much stitching touch-up time, then the system seems workable. Of course, if someone had to spend 6 hours on each of these examples, then it's a complete waste of time. I don't know anyone who owns the gigapan unit, and I know they just came out with the one made for DSLR's. I doubt my local rental house has one, and in any event that wouldn't get me access to the gigapan stitching software I don't think.

[Bart_van_der_Wolf]

I've had a crack at this ... thought it would be a great solution but....

1.  It is amazing how much depth of field is a problem.  Shooting at f16 would not get me sharp results shooting a 2m canvas from 5m! (without refocussing each image).  Refocussing each image is slow and tedious and introduces its own problems in stiching as the image magnification changes as you focus. Even if i did refocus for each shot, the frames towards the edge of the painting would suffer depth of field issues.

Hi Jeremy,

What focal length were you using? At 5 metres perpendicular (= 5.10 metres on the edge, or 5.20 metres in the extreme corner) a 100mm lens produces a DOF of 33 cm (= more than 12 inch) at f/7.1, and a 200mm lens still has 8.2 cm DOF at f/7.1 . That is calculated for a COC of 0.0096mm. An aperture f/16 would have increased that to 76 cm and 18.4 cm DOF respectively, but diffraction limited for 6.4 micron pitch sensor arrays.

Not refocusing is incorrect, a good pano sticher (PTAssembler, PTGUI) is able to compensate for the small differences in magnification by adding a focal length optimization (= magnification) into the equation.

2. If you are cross polarizing be aware that you will need to reset your camera polarizer as the camera moves. (This is not a problem if the lights are mounted to the camera platform and move with the camera) I ended up using a large sheet of polarizer on a c stand in front of the camera, not attached to the lens.

A good point, but I wonder how much effect the relatively small angles in your example had.

3. The stitching programs i tried were rarely able to automatically stitch the paintings together, forcing tedious control point marking for each image.

I don't know what kind of overlap you used, and whether you used a dedicated pano-head which allows to avoid parallax, but I usually have very little problems in getting a decent number of automatically determined control points, especially on flat sufaces without motion. Also the choice of a good pano-stitcher makes a lot of difference. PTAssembler even allows to stitch a flat plane when shot from different angles and not from a pano-head, by using its Camera Position Optimizer.

Cheers,
Bart

[JonathanBenoit]

You might not realize it, but this is disrespecting the artwork and the artist. All three examples are out of focus. I doubt you can even capture a sharp image with the gigapan.

Do you think there is no negative effect when you distort and stretch pixels?
This is frightening.

[Roscolo]

You might not realize it, but this is disrespecting the artwork and the artist. All three examples are out of focus. I doubt you can even capture a sharp image with the gigapan.

Do you think there is no negative effect when you distort and stretch pixels?

On my screen, when I zoom in, the examples not only are in focus, but are remarkably sharp. I can even count the canvas stitching. Are they perfectly sharp? Guess that depends on your definition of perfect. I'm not necessarily seeking perfection. There are certainly other ways to do this, as you posted above one way is to fork out $40,000. Frankly, $40K will buy a lot of 4x5 film and processing (already do my scans in-house). For my stated purpose of giclee' reproduction / documentation, if I can achieve the level of sharpness in the examples I posted, that level of sharpness is more than adequate for my purpose, and that is all that matters on this thread.

If I could achieve an acceptable result photographing large paintings with the gigapan, I can offer my existing clients a very, very good, although perhaps not "perfect" service for something they could not afford otherwise, as what I would have to charge to cover the $40,000 investment would be beyond the reach of most individuals' and institutions' budgets I currently shoot for. In other words, your $40,000 solution is not an option for me or my clients.

There is always a "negative effect" when reproducing artwork. It's never going to be a perfect match, whether you stretch pixels or just expose light to sensitized film or sensor. If it's possible to distort and stretch pixels to get a good result for my stated purpose and my artist or museum client is happy and within budget, I'm all for some distorting and stretching!

Every method of recording an image has advantages and disadvantages. The gigapan system may not work for 2-D art repro, but if I can duplicate the results in the examples from gigapan's website without too much TIME spent on image correction, then I think it is at least worth a little research into the possibilities.

[framah]

Regardless of what a poster said, the Betterlight system is still the best out there for accurate reproduction of art. I have just built a studio  with  moveable wall which allows me to shoot sections of the art with the lights and the camera remaining the same. It works great!

Kodak has stopped making the main component for the Betterlight scanback  and yes they have stopped production on 3 of their backs but the other three are still  available... all of which will give  you amazing resolution and a wide tonal range allowing you to print high quality repros.
Check out their website and decide for yourself.
Betterlight

[Roscolo]

Regardless of what a poster said, the Betterlight system is still the best out there for accurate reproduction of art. I have just built a studio  with  moveable wall which allows me to shoot sections of the art with the lights and the camera remaining the same. It works great!

Betterlight

And I would likely pick up a scanning back used or refurb-ed anyway. Some of my best equipment is no longer in production.

Congrats on the moveable wall, however, my museum / institutional clients aren't about to allow the works to be removed or transported anywhere off site. Work has to be shot on site, however, most have a space dedicated to viewing individual works that functions quite well as a studio. Should be hearing back from the gigapan folks in a day or two as to how well the system could be worked or tweaked to work with 2-d artwork documentation.

[BernardLanguillier]

Do you think there is no negative effect when you distort and stretch pixels?

There are valide reason to be concerned about the usage of stitching for fine art repro, but the key is to assess the actual impact of the various factors.

Issues with sharpness for those solutions unable to re-focus between frames seems to be a valid concern to me.

On the other hand, stitching has strong points also:
- virtually not light fall off at all which is a major advantage compared to 100% of the lenses out there,
- lack of distorsion of the final image

I have done pretty extensive comparisions between single shots from excellent lenses (like the Nikkor 24 PC-E) and stitches, and the stitch comes on top every single time as far as delivering an un-distorted image of the scene.

My personnal view is that the 3 key aspects for fine art repro are:
1. Accuracy of color
2. Accuracy of shape
3. Accuracy of brightness

To match the accuracy of shape delivered by stitching you will need to apply some level of impage manipulation to your single frame image, would it be 0.5% of distorsion correction, which will also stretch pixels.

This being said, there are very little measurable effects with such manipulations when they are done on healthy files like those of high end DSLRs. Sorry to say this, but my view is therefore that the concern you raise here is a philosophical one that is not backed up by actual facts.

Cheers,
Bernard

[Bart_van_der_Wolf]

Do you think there is no negative effect when you distort and stretch pixels?
This is frightening.

Hi Jonathan,

You seem to fear the resampling part very much. I wonder, why?

A good resampling method loses surprisingly little resolution, and what's lost can be mostly regained by restoration techniques like deconvolution sharpening. Only when extreme angles of view (e.g. wider than a 14mm single shot equivalent) are stitched together does one run the risk of interpolating to more than a single pixel in a diagonal direction (the tangential direction is still accurate to within a fraction of a pixel). The reproduction of most flat works of art is usually done with much narrower fields of view (also much easier to light).

Besides, it's very easy to super-sample the reproduction (just use a longer focal length and add more tiles), and then down-sample towards the desired output size. Proper down-sampling produces even better (anti-aliased) results than a direct single shot sensor recording.

Cheers,
Bart

[JonathanBenoit]

Hi Jonathan,

You seem to fear the resampling part very much. I wonder, why?

A good resampling method loses surprisingly little resolution, and what's lost can be mostly regained by restoration techniques like deconvolution sharpening. Only when extreme angles of view (e.g. wider than a 14mm single shot equivalent) are stitched together does one run the risk of interpolating to more than a single pixel in a diagonal direction (the tangential direction is still accurate to within a fraction of a pixel). The reproduction of most flat works of art is usually done with much narrower fields of view (also much easier to light).

Besides, it's very easy to super-sample the reproduction (just use a longer focal length and add more tiles), and then down-sample towards the desired output size. Proper down-sampling produces even better (anti-aliased) results than a direct single shot sensor recording.

Cheers,
Bart

As a technical photographer, I'm not interested in any loss in image quality. No loss is exceptable. I see a lot of people taking shortcuts. If you aren't going to have pride in your work, why do it at all.
Using stitched images that are not parallel to the material is the worst. There is no way to regain the correct shape of the artwork with any amount of post. There is a reason why view cameras are still highly used for technical photography. You may think you have distorted and stretched the pixels to match the material, but you have not and in the process you have horribly degraded the image.

You cant even compare the amount of detail from a one shot dslr to a MF back. The difference is exponentially more noticeable with a multishot digital back. This is one of those things that few have the opportunity to compare. Ive attached a .jpg comparison of the H 39mp single shot vs the multishot.

[Roscolo]

I don't want the thread to stray too far off topic, but since you use a photo of a US dollar, here is a link to an 8.2 Gigapixel image of a US 2 dollar bill made with the gigapan system I happened upon yesterday:

http://gigapan.org/gigapans/31604/

Looks pretty dang sharp to me when I can count the fibers in the paper. That said, I'm not looking to photograph dollars.

And for me, as I stated, if perfection doesn't fall into my budget or my clients' budgets, then if there is another way to achieve very, very good results, then that becomes acceptable for me. I'm not in the business of perfection for perfection's sake, I'm in the business of satisfying my clients. Maybe some of the museum jobs I've gotten in the past were because my fees were lower because I didn't have to factor in a $40,000 piece of equipment. I'm still a commercial photographer all these years, and I've watched a lot of friends go under because they overpaid or overborrowed to achieve "perfection." Folks like that usually mean good deals on some high-priced equipment for the rest of us at some point.

I respect your pursuit of perfection. I just cannot afford it. And moving forward in this thread I hope to hear from someone who maybe has used the gigapan system for the purpose of shooting paintings, or maybe someone who has the gigapan system will try it out on a piece of flat art and make a sample. I appreciate you don't find stitching acceptable. That's OK, but I know too many photogs who are stitching to produce hi-res giclee's to take your word over their outstanding results I've personally seen. They did invest a lot of time into stitching. Too much for me. I like the idea of the gigapan stitching system taking a lot of the work out of the equation and, as stated, if several hours of work is still required per image even using the gigapan system, then it's not workable for me and I would opt for a 4x5 scanning back as that comes closer to my budget, and my clients' budgets.

[JonathanBenoit]

I don't want the thread to stray too far off topic, but since you use a photo of a US dollar, here is a link to an 8.2 Gigapixel image of a US 2 dollar bill made with the gigapan system I happened upon yesterday:

http://gigapan.org/gigapans/31604/

Looks pretty dang sharp to me when I can count the fibers in the paper. That said, I'm not looking to photograph dollars.

And for me, as I stated, if perfection doesn't fall into my budget or my clients' budgets, then if there is another way to achieve very, very good results, then that becomes acceptable for me. I'm not in the business of perfection for perfection's sake, I'm in the business of satisfying my clients. Maybe some of the museum jobs I've gotten in the past were because my fees were lower because I didn't have to factor in a $40,000 piece of equipment. I'm still a commercial photographer all these years, and I've watched a lot of friends go under because they overpaid or overborrowed to achieve "perfection." Folks like that usually mean good deals on some high-priced equipment for the rest of us at some point.

I respect your pursuit of perfection. I just cannot afford it. And moving forward in this thread I hope to hear from someone who maybe has used the gigapan system for the purpose of shooting paintings, or maybe someone who has the gigapan system will try it out on a piece of flat art and make a sample. I appreciate you don't find stitching acceptable. That's OK, but I know too many photogs who are stitching to produce hi-res giclee's to take your word over their outstanding results I've personally seen. They did invest a lot of time into stitching. Too much for me. I like the idea of the gigapan stitching system taking a lot of the work out of the equation and, as stated, if several hours of work is still required per image even using the gigapan system, then it's not workable for me and I would opt for a 4x5 scanning back as that comes closer to my budget, and my clients' budgets.

I'm not against stitching. I'm against stitching images with a pano head, which from what I understand is exactly what the gigapan system is. The sensor has to be plume/parallel to the material or you have introduced some major skewing/distortion into the shape of the artwork

That two dollar bill is extremely blurry and out of focus.

I understand 40k is a lot to spend. Do whatever works for your business, just know that there is a huge difference in what you are producing with a gigapan and what the top photographers are with a true copy system. You can always rent the multishot back. I do work for a historic research library, so I know how it is. A day or two rental shouldn't be a problem. I think if you are going to short cut anyone, it should be the artists, not the museum.

[Roscolo]

That two dollar bill is extremely blurry and out of focus.

I think if you are going to short cut anyone, it should be the artists, not the museum.

Thanks for the input, but you lost some credibility with me with the criticism of the scanning backs for shooting artwork, one of the few uses for which a scanning back is ideal. And you mentioned the use of strobes. Since the advent of digital I abandoned strobes to photograph artwork years ago, and only use continuous lighting.

Let's just say that if I could achieve the level of "extreme blur and out of focus" you apparently see in the image of the 2 dollar bill, I and my clients will be very happy achieving such results of a painting with a sub $1,000 piece of equipment attached to the camera and stitching software.

Ironically, I'm interested in the gigapan because the museum is looking to save money anywhere they can. If I can save them money, I figure I can keep them as a client. Quality is important, but that quality has to fit into the client's budget or the customer is gone. They are looking to save money over film and scanning costs. The gigapan may be an option. Or a scanning back. I'm still intrigued by what may or may not be achievable with a gigapan.

Thanks for the input, but stitching, workarounds, and creative problem solving are a reality for most of us. By seeking alternatives and creative solutions and workarounds, not only am I not trying to "short cut" my clients, I'm putting in some extra time to try to solve their problem and find a solution they (and I) can AFFORD. No matter how good your solution is, if it's not affordable to the customer, it's meaningless. That's why this thread is called "gigapan system and shooting flat art."

Rental on the Hasselblad runs $450 a day, and that's not so bad (not great either) but the deposit is almost $30,000. What I would have to charge to not just cover expenses, but to cover the risk one takes with such a rental, is simply cost prohibitive to my clients. One problem I am not prepared to find a creative solution for is how to explain to my wife a $30,000 check or charge to my credit!

[BernardLanguillier]

Thanks for the input, but stitching, workarounds, and creative problem solving are a reality for most of us. By seeking alternatives and creative solutions and workarounds, not only am I not trying to "short cut" my clients, I'm putting in some extra time to try to solve their problem and find a solution they (and I) can AFFORD. No matter how good your solution is, if it's not affordable to the customer, it's meaningless. That's why this thread is called "gigapan system and shooting flat art."

Speaking about stitching, you might want to consider these options as well:

http://www.dr-clauss.de/VRStationHD_EN.htm

IMHO much better than gigapan, but also more expensive obviously. Still reasonnable for pro use I believe. Autopano pro support their XML files which should ensure very accurate stitching even when there is too little texture in some parts of a piece.

They have also lower spec version, still way more robust than the gigapan:

http://www.dr-clauss.de/VRStationST_EN.htm

Regards,
Bernard

[Thomas Krüger]

The stitching programs i tried were rarely able to automatically stitch the paintings together, forcing tedious control point marking for each image.

From the PTGui FAQ's: How can I stitch mosaics, like partial scans from a flatbad scanner of a large image?
http://www.ptgui.com/support.html#6_6

[JonathanBenoit]

Thanks for the input, but you lost some credibility with me with the criticism of the scanning backs for shooting artwork, one of the few uses for which a scanning back is ideal. And you mentioned the use of strobes. Since the advent of digital I abandoned strobes to photograph artwork years ago, and only use continuous lighting.

Let's just say that if I could achieve the level of "extreme blur and out of focus" you apparently see in the image of the 2 dollar bill, I and my clients will be very happy achieving such results of a painting with a sub $1,000 piece of equipment attached to the camera and stitching software.

Ironically, I'm interested in the gigapan because the museum is looking to save money anywhere they can. If I can save them money, I figure I can keep them as a client. Quality is important, but that quality has to fit into the client's budget or the customer is gone. They are looking to save money over film and scanning costs. The gigapan may be an option. Or a scanning back. I'm still intrigued by what may or may not be achievable with a gigapan.

Thanks for the input, but stitching, workarounds, and creative problem solving are a reality for most of us. By seeking alternatives and creative solutions and workarounds, not only am I not trying to "short cut" my clients, I'm putting in some extra time to try to solve their problem and find a solution they (and I) can AFFORD. No matter how good your solution is, if it's not affordable to the customer, it's meaningless. That's why this thread is called "gigapan system and shooting flat art."

Rental on the Hasselblad runs $450 a day, and that's not so bad (not great either) but the deposit is almost $30,000. What I would have to charge to not just cover expenses, but to cover the risk one takes with such a rental, is simply cost prohibitive to my clients. One problem I am not prepared to find a creative solution for is how to explain to my wife a $30,000 check or charge to my credit!

Benefits of strobe is that they are daylight color accurate. Potentially you could use HID lighting, but I've read some weird information on how you need to leave them on for extended time. Something to do with the gases/filament. I cant remember. I am wondering if you are using hot lights or fluorescents. I hope not, but it wouldnt surprise me if you are using hazardous lighting. What are you using for lighting?

Scanning backs are a good option, but they don't have the same level of detail that the mutilshot backs have. Especially when coupled with proper software like Phocus or CaptureShop. They are also slowly dieing off. I cant see myself investing in a scanning back when no one will be making them in a year. I was informed by Better Light that they are selling off their remaining inventory and only servicing backs in the future. That's not a comfortable investment. Although, it's still a much better idea than the gigapan.

Another option is to use your current camera and purchase a sliding table made by Velmex. You can even get it motorized. The table will slide while keeping the material flat allowing you to capture sections of the art while keeping the art parallel to the sensor plane. You can then properly stitch the material in PS.

You should have equipment insurance along with general liability. Mine is about $600 a year through APA insurance. Gives you peace of mind on your equipment and this way you don't need a deposit when renting a digital back or any other rentals.