Cross-polarized light setup for photographing artwork

[John V.]

I have a budget setup so I can photograph a personal collection of artwork for reproduction, and it pretty much goes like this:

The painting is stationary on a wall and the camera is on a slider with adjustable height. (I shoot in sections and then stitch together)
2 umbrella lights with polarizing filters @45 degrees or  so, polarizing lens filter for the camera.

New to this type of setup... (some are oil paintings, reflecting a lot of light)

Looking for a little help...

1) Does the light have to be directly hitting the painting, and not bouncing off the umbrellas?

2) Does cross polarization work with the camera moving to multiple positions?

What type of light are best for photographing artworks?

Any help or advice is appreciated, thank you.

[John Nollendorfs]

By bouncing the polarized light into the umbrella, you are effectively negating it! Most people will use rather small polarized light sources and move them as far as possible from the paintings to minimize specular reflections and make the lighting as even as possible.

[John V.]

I figured this was the case regarding that issue! It's difficult getting a straight answer out of the internet sometimes.

However, I am still wondering.

All I have to do is flood the artwork with even lighting(through the polarizing filter only), and I can move the camera side to side and up and down and I will still achieve the same polarizing effect? As in, The camera does not have to be "aligned" with anything? ( I've seen numerous tutorials on how to photography artwork with a cross polarization setup *while the camera is stationary*) Pardon my simplicity

[D Fosse]

It's difficult getting a straight answer out of the internet sometimes.

Well, you just got one there. No umbrellas, you want the light source to be as small as possible. You pretty much have to use strobes (and no modeling light), polarizing film is usually not heat resistant.

I wouldn't recommend moving the camera, it's easy to get into alignment trouble. Instead use a long lens, and rotate the camera around the optical center of the lens. That will be where the diaphragm appears to be (which isn't where it actually is) when you look into it from the front. This way you eliminate parallax error. There are adjustable rails available for this (a little experimenting will be necessary to determine the exact position). A more specialized solution is the Gigapan, which is programmable to shoot at regular intervals.

You will probably have to correct geometry in Photoshop afterwards. Hard to avoid when stitching. It may help to shoot a single-frame "guide shot" and upsample that to desired size.

Oh, and if you process in Lr/ACR - apply lens correction before stitching.

[elf]

I think you also need to adjust the polarization at the lighting, not at the camera.

[D Fosse]

It's possible that it is slightly more or less effective depending on the angle it's mounted on the lights. I don't know, I have never noticed any difference. As long as they are oriented at 90 degrees to each other, it's extremely effective in any case. This is easy to check:

[LKaven]

Lights can be two strobes in bowl reflector angled in at 45 degrees, each with a sheet of Rosco polarizing gel.  Orient the gel the same way on both lights (e.g., vertical).  Using the modeling lights (temporarily) you can dial in the correct orientation for the circular polarizer while looking through the viewfinder.  Watch for the specular reflections to disappear. 

Good lens for it, if for example using a Nikon, would be the 60mm f/2.8 AF-s micro.  The 85/1.8g is also good.

The only difficulty with the polarizing method is that it can induce color shift.  You'd want to have a color reference handy, e.g., the QPCard, or the Xrite Passport.   

[D Fosse]

Good lens for it, if for example using a Nikon, would be the 60mm f/2.8 AF-s micro.  The 85/1.8g is also good.

Yes, those two are my favorites for art repro. The micro 60/2.8 is crazy sharp, flat field and with very little distortion. I've always felt it's underrated.

It's a bit short for stitching, though, even with a DX sensor. The 85/1.8, also a great lens, would be minimum IMO, but depending on the number of individual segments I'd go up to 135 - 200. The total merged result should correspond to a standard angle of view.

[adpix]

A brief note on umbrellas:  while bouncing light from an umbrella will negate polarization, you can shoot through an umbrella or other scrim without affecting polarization. I do it all the time in shooting art that incorporates metallic elements, so as to reduce speculate  highlights while avoiding the "angle of incidence equals angle of reflection" effect, which can leave parts of a metal surface unlit and dark.

[adpix]

That should read "specular highlights."

[Jason DiMichele]

It may help to shoot a single-frame "guide shot" and upsample that to desired size.

That is a really good tip. Thanks for that!

[Jason DiMichele]

I'm currently using two strip soft boxes for even soft light with a sheet of polarizing film over each. I then have a polarizing filter on the lens. I typically stitch with a 100mm macro or 200mm lens and autopano giga (pro will work fine) for stitching. I also create a custom camera profile under the polarized light with a passport colour checker. Since most of the time the polarization is consistent between repros for paintings the camera profiling works really well.

[D Fosse]

That is a really good tip. Thanks for that!

Yeah, I work at an art museum and used to stitch all the time, working with a D700. I would of course never have enough resolution, so it became a matter of some urgency to streamline the process.

So I found that if I set up the shot with, say, a 50mm, I could just switch to a longer lens without moving anything in the setup, and do 6-10 segments. The segments will then align perfectly with an upsampled guide shot used as a base for the merge (since the total angle of view is the same), and it also eliminates the need for further geometry adjustment. The trick is to upsample to a slightly smaller size than the merge would normally end up at, so that the merge process only samples down, not up.

The rotation point has to be set up for the long lens in advance.

And then of course I got the D800 and all of this was no longer needed. The results I get now are just as crisp at the same resolution, as long as I use live view focusing, mirror up and all that.

Using strip soft boxes, or even umbrellas, together with polarizing film is something I haven't tried, but it sounds like it's worth some experimenting. The "dead metal" effect (particularly in frames) I have just taken for granted as an unavoidable side effect.

[Jason DiMichele]

That's a very interesting tip about the technique to setup the shot for stitching. I typically would just move back as far as I need to in order to have the overlap required for the resolution I wanted. Your method makes a lot of sense and I will try that. Instead of moving the camera back I will just choose the appropriate focal length that will provide me the desired final resolution.  I'm using a custom build spherical pano head that I have calibrated to the lenses I use.

So you are finding that since most paintings don't have an incredible amount of fine detail (photographically speaking) that 36mp is enough to enlarge to virtually any sized print or just to a print the same size as the original painting? I'm using a Canon 5D MkII at 21mp. I usually stitch about 6-10 images.

Umbrellas may not work with the polarizing film because I've only seen it sold in about 18" strips so that's why I went with the strip softboxes.

Another question I have for you is how you hold the art while reproducing. I currently hang it on the wall with plastic spacers on the bottom to keep the painting flush. I'm thinking about moving to an easel setup with the only caveat being that I have to ensure that the camera is totally parallel to the painting.

[adpix]

No, no. You don't put the polarizing film over the umbrella or soft box.  Rather, you attach a standard reflector to your strobe. Next you use a clip to attach the polarizing filter and such gels as you need for color balance, to the reflector. Then you attach an umbrella so that the shaft pokes through the gels (cut a small slit) and then the shaft locks into the strobe head. Very easy and very effective.

As a footnote, you need a color meter to verify that both strobe heads have the same Kelvin temp. Flash tubes vary. You correct the variance with Rosco-like gels placed between the flash tube and the polarizing gel.

[Jason DiMichele]

I am currently putting the polarizing film over my soft boxes with Velcro. I get a softer, more diffused light. That seems quite a bit easier and quicker than your umbrella method with the added advantage of the soft boxes giving more focused light than an umbrella. I am also pretty sure that when polarized light is bounced or shot through something (diffusing panel, umbrella, etc) it loses it's polarization. Not sure why I would need to colour balance anything since the strobe is daylight balanced. I make custom camera profiles either way.

I'm using Profoto lights. Probably as close as you're gonna get in terms of colour consistency between lights that the human eye would notice. I suppose I could shoot a white board to verify colour accuracy but I can't say I've noticed any difference with light temperature.

Have you tried both methods and seen better results or is it just a different approach you prefer using?

[D Fosse]

Another question I have for you is how you hold the art while reproducing. I currently hang it on the wall with plastic spacers on the bottom to keep the painting flush. I'm thinking about moving to an easel setup with the only caveat being that I have to ensure that the camera is totally parallel to the painting.

For smaller works I use a large repro stand, and place a mirror on the base. When the lens is perfectly centered in the viewfinder it's parallell.

For larger paintings I use an easel, but you're right, ensuring camera and painting are perfectly parallell is a challenge. I'm working on that, hoping to find a way to use the mirror method there as well. Any suggestions?

Assuming the frame is perfectly rectangular is of course not a good assumption, and it doesn't take much to throw a corner slightly out of focus.

[D Fosse]

So you are finding that since most paintings don't have an incredible amount of fine detail (photographically speaking) that 36mp is enough to enlarge to virtually any sized print or just to a print the same size as the original painting? I'm using a Canon 5D MkII at 21mp. I usually stitch about 6-10 images.

Yes, for most practical purposes. But photographic technique is all - heavy, sturdy tripod, maximum enlarged live view focusing, mirror up, cable release or timer. And of course only the sharpest possible lenses, stopped down to 5.6 or 8, never more. The Micro 60/2.8 and the 85/1.8 are my current workhorses. These lenses really resolve down to pixel level with the D800.

I do occasional stitching with the D800, but only in very special cases, if the work is extremely large and fine detail is still necessary. But normally a 36 MP original goes down to individual fibers in the canvas, sharply resolved. There's really nothing useful beyond that.

[adpix]

One reason not to cover your soft box with a polarizing filter is simple cost. A smaller filter over the reflector is much less expensive. That is especially revelant to my practice of shooting through two six by six foot scrims.

The fact that polarization is not lost through scrims is easily demonstrated. Set up your lights as usual, turn on the modeling lights on the strobes and kill the room lights. Then take a bottle of colored glass and invert it on a light stand in front of your art. Look through the lens and will see two reflections on the bottle. If your polarizing filters on the lights and camera are properly aligned, you will see the two reflections disappear as you rotate the polarizing filter on your lens. I use this method all the time to establish the desired degree of polarization.

I can't speak to Profotos, but most studio strobes that I've used have a variance in Kelvin temp from head to head. You need a color meter to detect it and correct with gels. Plus I have found that keeping the over all color temp in the 5000-5500 range gives me a better image, as opposed shooting too warm or cold, and then correcting in post processing.

[D Fosse]

The fact that polarization is not lost through scrims

This is interesting. So you're saying I can use polarizer with, say, a large softbox with full effect? I always just assumed off-hand that wouldn't work.

Experimenting time ahead. The problem is to find a way to attach it inside the softbox, close, but not too close, to the tubes (too close, and the film will burn out and curl). There's no obvious place with my Elinchroms, but I'll think of something.