How to Photograph a Card So Its Centering Measures Correctly
A repeatable protocol for photographing cards to measure centering: why tilt is the biggest error, the geometry with worked numbers, lighting, and the 180 test.
A centering measurement taken from a photo is only as good as the photo. The arithmetic is trivial and exact; every meaningful error is baked into the image before any software sees it. The single biggest one is perspective: if your phone is not parallel to the card, the border nearer the lens is magnified relative to the one further away, and the measurement drifts by more than the gap between a PSA 10 and a PSA 9.
The good news is that the fix is mechanical rather than skilful. Card flat, camera straight down, two diffuse lights, card fills the frame, shoot the back too, then rotate the card 180 degrees and shoot again. That protocol gets a phone photo to roughly the same precision as a flatbed scan. This article explains why each step is there, with the geometry worked out so you can see how much each mistake costs.
Why perspective is the whole game
Model the camera as a pinhole. An object's size in the image scales as 1 divided by its distance from the lens. If the card is exactly parallel to the sensor, every point on it is the same distance away, so the whole card is scaled by one number and all the border ratios survive untouched. Tilt the card (or the phone) and that stops being true.
Take a standard card, 63.5 x 88.9 mm. Tilt it by an angle about its horizontal centreline, so the top edge rises toward the lens and the bottom edge falls away. The top and bottom edges are each half the card height from the centre, so they move toward and away from the lens by 44.45 x sin(angle).
At a camera height of 200 mm above the card centre:
| Tilt | Edge moves by | Near edge distance | Far edge distance | Magnification ratio near : far | A true 50/50 pair reads as |
|---|---|---|---|---|---|
| 2 degrees | 1.55 mm | 198.4 mm | 201.6 mm | 1.016 | 50.4/49.6 |
| 5 degrees | 3.87 mm | 196.1 mm | 203.9 mm | 1.040 | 51.0/49.0 |
| 10 degrees | 7.72 mm | 192.3 mm | 207.7 mm | 1.080 | 51.9/48.1 |
| 15 degrees | 11.51 mm | 188.5 mm | 211.5 mm | 1.122 | 52.9/47.1 |
Working the 10 degree row through explicitly: two genuinely equal 2.00 mm borders are imaged at sizes in the ratio 207.7 : 192.3, which is 1.080 : 1. Scaled so the far border reads 2.00 mm, the near one reads 2.16 mm. The measured ratio is 2.16 / (2.16 + 2.00) = 51.9%, so 52/48 on a card that is perfectly centred.
Two things follow.
It compounds with the real shift. Perspective does not just add noise; it has a direction. If the wide border happens to be on the near side, the error adds to it. A card genuinely at 55/45 photographed with a 10 degree tilt in the unlucky direction reads around 57/43, pushing a card that sits on PSA's published Gem line into the zone where it plainly does not. In the lucky direction, the same tilt hides a real problem.
Distance matters as much as angle. The same 10 degree tilt from 150 mm gives near and far distances of about 142 mm and 158 mm, a ratio of 1.11, which reads as roughly 52.6/47.4. From 400 mm the same tilt gives 392 mm and 408 mm, a ratio of 1.04, or about 51/49. Shooting from further back with a longer lens roughly halves the penalty for the same sloppiness.
Assumptions in all of this: an ideal pinhole projection, a perfectly flat card, tilt about the long axis, and distances measured to the card centre. Real lenses add their own distortion on top, which is the next section. The point is not the exact decimal, it is the order of magnitude: a tilt you cannot see by eye moves the answer by about as much as the difference between adjacent grade lines.
Being off-centre in the frame is fine; being tilted is not
This surprises people. If the phone is genuinely parallel to the card, it does not matter where in the frame the card sits. A flat rectangle parallel to the sensor projects to a scaled copy of itself regardless of position, so the border ratios are preserved exactly.
What breaks that is lens distortion, which is a property of the glass rather than the geometry. Barrel distortion is strongest away from the optical axis, and phone ultra-wide lenses have a lot of it. A card sitting in the corner of an ultra-wide frame can have its outer edge visibly bowed, and a bowed edge is an unmeasurable edge.
So the practical rule is: centre the card because it costs nothing and removes distortion from the equation, but spend your attention on getting the phone parallel, which is the error that actually scales with your carelessness.
The protocol
1. Take the card out of the sleeve. A penny sleeve's edge sits roughly a millimetre outside the card edge and is frequently the crispest line in the frame, which is exactly what edge detection is looking for. Toploaders add glare, a bevel and a second outline. Every grading company measures the bare card; so should you. Handle by the edges and re-sleeve as soon as you are done. Measuring through plastic when you have no choice is covered in measuring centering through sleeves, toploaders and slabs.
2. Lay it flat on a matte, contrasting background. Mid-grey works for almost everything. A white background under a white-bordered sports card makes the real edge disappear; a black background under a dark-bordered card does the same. Matte matters as much as the colour, because a glossy surface reflects the card and gives you a phantom second edge.
3. Flatten a curled card first. Holos and refractors often bow. A bowed card lifts one edge toward the lens, which is the tilt problem in miniature and applies to one side only. Let the card sit flat for a while under something clean and light; do not press or bend it.
4. Set up two diffuse light sources at about 45 degrees, from opposite sides. Single-sided light throws a thin shadow along the far edge of the card. On a dark background that shadow merges with the card and pushes the detected edge outward; a 0.2 mm shadow read as border on a 2 mm border turns 50/50 into about 52.4/47.6. Two opposed sources cancel each other's shadows. A window on an overcast day plus a lamp bounced off a white wall is perfectly adequate.
5. No flash, ever. An on-camera flash sits next to the lens, so it reflects straight back off any gloss and blows out a patch of the card, often across a border. Holo and foil cards are the worst case.
6. Get the phone parallel. This is the step that matters most. Use the built-in level (many camera apps show a crosshair or horizon indicator when you point straight down), or the grid overlay, or rest the phone on a shelf, a stack of books or a cheap overhead clamp. If you are hand-holding, brace your elbows and take three shots.
7. Shoot from further back with the main or telephoto lens. Avoid the ultra-wide. Aim for 200 to 400 mm from the card if your lens will focus there.
8. Fill the frame with the card. This is where your resolution comes from. If the card's 88.9 mm height spans about 2,800 pixels, you have roughly 31 pixels per millimetre, or 0.032 mm per pixel, which is finer than a 600 dpi scan at 0.042 mm per pixel. If the card fills a third of the frame you have thrown two thirds of that away, and digital zoom does not give it back.
9. Lock focus and exposure. Tap the card, then hold to lock. A soft edge is an ambiguous edge, and autoexposure hunting between shots changes where the border appears to end.
10. Shoot the back the same way. PSA, BGS, CGC and TAG all publish a back tolerance (SGC is the exception and states none), and backs are routinely worse than fronts because the two faces are printed and registered separately. A card can lose a top grade on the back alone.
The 180 degree test
This is the most useful habit in the whole protocol, and almost nobody does it.
After you have measured the card, rotate it 180 degrees in the same setup, without moving the camera or the lights, and measure again.
- A real off-centre shift belongs to the card. Rotate the card and the wide border stays on the same side of the card, so it appears on the opposite side of the frame.
- A perspective or lighting error belongs to the setup. Rotate the card and the error stays on the same side of the frame.
So compare the two readings. If the first shot says 57/43 with the wide border on the left of the card, and the rotated shot says 57/43 with the wide border still on the left of the card, the card really is 57/43 and your setup is clean. If the first says 57/43 and the rotated one says 51/49, your camera was tilted and the truth is somewhere near the average, roughly 54/46, with error bars you should not ignore.
Averaging the two readings actually cancels first-order tilt error, which is why this trick is worth the extra thirty seconds on any card you are deciding to submit.
Measure each side at two points
While you are at it: measure each border near both ends, about 15 mm in from each corner, rather than once in the middle.
A card cut slightly out of square has a border that tapers, wider at one end of a side than the other. Collectors call it a diamond cut or a tilted cut, and the grading companies name it explicitly. Beckett's scale allows "very slight diamond cutting" at BGS 7 and progressively more at lower grades, down to "heavy diamond cutting" at BGS 1; SGC notes at grade 7 that a card may exhibit a slightly skewed (diamond) cut.
If you take one reading per side, a tapered card measures whatever you happened to point at. Two readings per side, averaged, give you a fair ratio and also tell you the card is skewed, which affects eye appeal even when the average passes. A difference of more than about 0.2 mm between the two ends of one side is worth noting.
From photo to number
Once the image is right, the measurement is mechanical. Left percent = left border divided by (left plus right), times 100; the same for top and bottom; write the larger number first. Measure from the physical card edge to the outer edge of the printed border, never to the artwork, because the artwork is placed by the designer and tells you nothing about where the cut fell.
You can count pixels in any image editor, or place the guide lines yourself in the free card centering calculator, which does the arithmetic and compares the result to the published PSA, BGS, CGC, SGC and TAG lines. The Card Centering Calculator app measures the same thing from a photo and maps it to the same standards. Either way the software is doing what you would do by hand; the reason to use it is consistency, not magic. The full set of methods, including calipers and pixel counting, is in how to measure card centering at home, and an honest account of how much error each tool carries is in do card centering apps and tools actually work?.
Then read the number against the standard you care about. PSA's published Gem Mint 10 line is approximately 55/45 on the front and 75/25 on the back, with 60/40 and 90/10 at 9. PSA states that centering "is designated as the percent of difference at the most off-center part of the card," so the worse of your two axes is the one that counts, on both faces.
When to use a scanner instead
A flatbed scanner has no tilt and no perspective, because it images a line at a time at a fixed geometry, and its resolution is known exactly: at 600 dpi one pixel is 25.4 / 600 = 0.042 mm, at 1200 dpi it is 0.021 mm. If you own one, it is the more repeatable instrument, and it removes the single biggest error source in this article by construction.
A phone wins on speed, on surface and gloss inspection under raking light, and on cards you cannot press under a lid. A scanner wins on repeatability. The trade-offs are worked through in flatbed scanner vs phone camera for measuring centering.
Common mistakes
- Hand-holding at an angle. A tilt too small to notice moves the ratio by one to three points.
- Shooting through a sleeve or toploader. The sleeve edge becomes the measured edge.
- One light source. The shadow on the far edge becomes border.
- Flash on a holo. The blown-out patch erases the border line exactly where you need it.
- Card small in the frame. You throw away the resolution that makes a phone competitive with a scan.
- Ultra-wide lens up close. Maximum perspective and maximum barrel distortion at once.
- Only shooting the front. Backs have their own published tolerance at PSA, BGS, CGC and TAG, and are usually worse.
- One reading per side. Misses tapered and diamond cuts entirely.
- Skipping the 180 degree check. It is the only way to tell a real shift from your own camera angle without a second instrument.
Bottom line
- Perspective is the dominant error. A 10 degree tilt from 200 mm reads a perfectly centred card as about 52/48; from 150 mm it reads about 52.6/47.4. That is the same size as the gap between adjacent grade lines.
- Position in the frame barely matters if the phone is parallel; lens distortion is the only reason to centre the card, and it is worst on ultra-wides.
- Card out of the sleeve, flat, matte contrasting background, two diffuse lights at 45 degrees from opposite sides, no flash, phone parallel and straight down, card filling the frame, focus and exposure locked, back shot too.
- Fill the frame and a phone gives about 0.03 mm per pixel, finer than a 600 dpi scan at 0.042 mm.
- Rotate the card 180 degrees and re-measure. A real shift follows the card; a camera error follows the frame. Averaging the two cancels most tilt error.
- Measure each side at two points about 15 mm from the corners to catch tapered and diamond cuts, and always measure the back as well as the front.
Frequently asked questions
How should I photograph a card to measure its centering?
Take the card out of its sleeve, lay it flat on a matte surface that contrasts with the border, and shoot straight down from directly above the centre of the card with the phone parallel to the table. Light it with two diffuse sources at about 45 degrees from opposite sides, no flash. Fill the frame with the card, tap to lock focus and exposure, and shoot the back the same way. Then rotate the card 180 degrees and shoot again as a check.
Why does a tilted photo change the centering measurement?
Because the edge of the card nearer the lens is magnified relative to the edge further away. With a standard 88.9 mm tall card photographed from 200 mm, a 10 degree tilt puts the near edge about 192 mm from the lens and the far edge about 208 mm, so the near border is magnified by roughly 1.08 relative to the far one. Two genuinely equal 2.0 mm borders then measure about 2.16 mm and 2.00 mm, which reads as 52/48 on a card that is really 50/50.
Does it matter if the card is off to the side of the frame?
Less than you would think, as long as the phone is parallel to the card. When the sensor plane and the card are parallel, the image of a flat rectangle is a pure uniform scaling no matter where in the frame it sits, so the border ratios are preserved. What does hurt is lens distortion, which is strongest at the edges of the frame and worst on ultra-wide lenses. Centre the card anyway; it costs nothing and removes the issue.
Should I photograph the card in its sleeve?
No. A penny sleeve's edge sits outside the card edge by around a millimetre and is often the sharpest line in the picture, so automatic edge detection can lock onto the sleeve instead of the card. Plastic also adds glare and a second set of edges. Graders measure the bare card, so your photo should show the bare card. Take it out, shoot it, and re-sleeve it immediately.
How do I check whether my photo measurement is trustworthy?
Rotate the card 180 degrees in the same setup and measure again. A genuine off-centre shift belongs to the card, so the wide border stays on the same side of the card. A perspective error belongs to the camera, so the wide border stays on the same side of the frame. If the two measurements agree within a point or so, your setup is sound. If they disagree by several points, the camera was tilted and neither number is usable.
Which phone lens should I use for card photos?
The main 1x lens or a 2x telephoto, held further back, rather than the ultra-wide. A wide lens close to the card exaggerates every perspective effect and adds the most barrel distortion at the frame edges. Backing off and using a longer lens flattens the geometry, and modern phones have enough resolution that a card filling most of the frame still gives you around 30 pixels per millimetre, which is finer than a 600 dpi scan.
