Coin Photography is not easy

[ray_parkhurst]

Hi @rbenton,

I do see how shooting a white background would help with processing, and agree that a black background would allow a similar process, but doing the processing at all should be avoided IMO. Part of the reason I do "disappearing" backgrounds is to avoid such processing. You could also avoid it with the white background, simply doing a manual crop around the coin. Allowing Photoshop to mess with it creates an unnatural edge treatment. I've never done a post-processed black background on a coin that was shot on black, so am not sure it is any better, but I would expect so, since the transition from the darker coin edges to the background is a more natural one versus the dark edges to white. Publishing in black may mitigate this, even if you shoot backlit.

As examples, your 1915 edges with white background look very processed, but the 1955 edges do not, see below. However, your 1955 image is titled "55 ddo phone", and is lower quality vs the 1915 image. Was it shot and processed the same way? Can you process the 1915 (backlit) image with a final black background for comparison?

And not to be nitpicky, but your 1915 image seems a bit oversharpened. Sharpening level is a personal choice of course, and the images I am showing below are at 300% so suffer from pixelation, but at these zoom levels it's usually more apparent if oversharpening has occurred. You didn't mention how much sharpening you're using, or how it's being done. I see this type of oversharpening happen with Sony cameras (using Sony software), but I don't have any experience with the D810 and PhotoShop. Using LightRoom, I can create such adverse effects most easily using the Local Contrast adjustments, or with the Contrast by Detail Level adjustments.

If you want to try shooting with a black background, I use ProtoStar. It is the least reflective surface I have found which can stand up to background use, ie having a coin sit directly on it without damaging it. There are other materials and paints and such that are much blacker, but they are too delicate for such use. Even the ProtoStar material must have some maintenance to remove any dust before shooting, but it holds up well.

One more thing about the metadata...I find it odd that you're not doing MWB. How can you trust the color of your output?




.
.
.
Edited to add: I see a couple other items in the metadata to discuss:

ISO 320. Because you're shooting hot, this works OK, but if you start shooting to avoid over-exposure, you will want to go lower in ISO (100 or lower) to improve shadow noise. This is especially important as you map shadows from raw to jpg. Note that if you use a black background like the ProtoStar, you'll lose some shadow detail unless you change the processing a bit since even the blackest backgrounds don't go to 0,0,0. I believe what you're doing with the backlighting and over-exposure is called "shooting to the right", while what you end up doing with black backgrounds is "shooting to the left".

Aperture 16. I assume this is the readout from your AFS lens, which will report effective aperture rather than nominal. However, EA16 is still quite small, and while it improves DOF, it also degrades resolution and sharpness. Your D810 pixel pitch is 4.88um, so your DLA is ~f/7. You say D810, not D810E, so you have an active AA filter that will degrade the DLA by perhaps 50% to ~f10, so you are losing 60% of your available resolution and sharpness by shooting at EA16.

[rbenton]

My method while batch processing does not cut/paste the coin image. The steps are: open the image, select the coin, crop tightly to that selection, then (after deselecting) enlarge the canvas size by 104% to add some white space around the tightly cropped image of the coin. The result is an image of the coin with backlit background, and some digitally added white space around it to give the coin some 'breathing room.' I only use this approach when I am dealing with dozens of coin images at once. It is much cleaner, more consistent, and faster than the way I would do a single coin. If I only have one coin to shoot, I use any blank surface as a background. This of course produces a visible cast shadow of the coin. In post I would then manually select the coin without including the shadow, then cut/paste into a blank white frame. That is what I did with the 1915 image and the artifacts you see around the edge are vestiges of the original background around an imperfect selection. The 1955 w/black background was done this way, but any remnants of background color that may have remained disappear into the blackness.

I add some sharpening in Photoshop's Camera Raw where there is a group of sliders to adjust values for Sharpening, Pixel Radius, Detail, and Masking. Sharpening here is much more limited than other methods in Photoshop, such as Unsharp Masking, or High Pass sharpening which can easily be overdone. I use 75% sharpening at 1.5 pixel radius and leave the other sliders alone.

When prepare to shoot a large number of coins, I take care to keep the plane of the camera sensor parallel to the plane of the coin, I manually focus using the live view zoomed in as far as possible then I lock the focus ring in place with gaffer tape. I adjust the backlight until it's just barely overexposing. Finally I shoot a grey target which I use to establish WB in post. I apply that value to all the coin images. If I am just shooting one coin, it's much more 'quick-and-dirty.' I hand-hold the camera and set WB for daylight which is usually close enough.

Finally, I think your advice about pixel pitch and DLA is very good. I will put that to use. Do you have a good resource for calculating this? I have a couple other cameras I use and the numbers would certainly be different if I shoot macro with them (Nikon D850 and D4S).

Here is a coin I shot today using all the 'batch' methods mentioned. So except for the 4% white border, the background is in-camera. The edges of the coin are complete and the color and exposure are as close to 'reality' as I could manage. (I used a 18% grey card for exposure and WB)

Thanks again, Ray. I love the learning process!

[ray_parkhurst]

When you increase the canvas size, do you scale the coin image? Non-integer scaling degrades sharpness, and sub-fractional scaling is the worst offender. I saw the "104%" and cringed, but if you're just adding space around the coin and not affecting the coin image itself, no worries.

I forgot to comment before regarding the image rotation. It's best to avoid this as it does degrade sharpness, though not as much as non-integer or sub-fractional scaling does.

As I said before, sharpening is personal and subjective, and there is a wide range of sharpening in various coin images on the web. I only commented because the result I see on your images was similar (though not as dramatic) vs what I saw on the A7Rm3 and A7Rm4 with pixel-shift.

Given your process, I would not see any advantage to using a wider aperture. Shooting at EA16 gives you enough DOF to compensate for the variations in critical focus distance when using fixed focus. Ensuring flatness is important for either fixed focus at EA16, or critical focusing at wider aperture, so is always a good practice. If you used a wider aperture you'd need to critically focus each shot at a mid-height feature.

Back to image rotation, I see that you did an ~45deg rotation CCW on this coin. This, plus your canvas expansion and the not-quite saturated backlighting, leaves some interesting background artifacts, see below. If I were using such a process, I'd further increase the backlight brightness to ensure background saturation, and adjust my direct lighting to compensate. I think this would eliminate some of the issues with artifacts and may possibly simplify the workflow as there would be no need to crop the coin and paste white around it...it would already be there. If the camera does not respond well, and does not push the background to 255,255,255, then a simple brightness or exposure adjustment in post should give the same result. I was pleased to see no over-exposure in the '09-S image, but you may need to do more exposure compensation if you increase backlight brightness to give room for the added post processing.

Thanks for the explanation of how you're doing white balance. I will often shoot with a grey background, especially with toned coins, to ensure I carry a reference in the image. Makes people believe the colors.