Tools & Apps12 min read

Flatbed Scanner vs Phone Camera for Measuring Card Centering

Scanner or phone for measuring card centering? The dpi arithmetic, why a scan has no perspective error, Newton rings and lid pressure, and which to use when.

By

A sleeved trading card and a jeweler's loupe on a light grey desk

A flatbed scanner beats a phone camera for measuring centering, and the reason has almost nothing to do with megapixels. A scan is taken with the sensor a fixed distance from a flat glass plate, moving parallel to it, so there is no perspective error and the scale is known before you start: set 600 dpi and every pixel in the file is exactly 25.4 / 600 = 0.0423 mm wide. A phone has to be held parallel to the card by hand, and a tilt of a few degrees is enough to move the measured ratio by a point or two, which is a meaningful fraction of the gap between PSA's published Gem Mint line of approximately 55/45 and its Mint 9 line of approximately 60/40.

That said, a modern phone has more than enough resolution, it is faster, and it shows gloss, holo pattern and surface defects that a scanner flattens out. The right answer is usually scan when the decision costs money, photograph when you are screening.

Below: the dpi arithmetic worked out in full with its assumptions stated, what resolution actually buys you, the scanner problems nobody warns you about, the phone problems you can fix, and a straight recommendation per use case.

The arithmetic: turning dpi into millimeters per pixel

Everything here rests on one exact conversion: one inch is 25.4 millimeters, by definition. So:

mm per pixel = 25.4 ÷ dpi

Four assumptions before the numbers, because they are where people go wrong:

  1. The dpi is the scanner's optical resolution, not an interpolated setting. Software will happily offer a resolution above what the sensor and stepper motor can actually resolve. Interpolated pixels are invented by resampling; they make the file bigger and the measurement no better. Use your scanner's native optical figure.
  2. You are counting pixels in the unresampled scan at 100% zoom. If your viewer is showing the image scaled to fit, you are reading screen pixels, not image pixels.
  3. The card lies flat on the glass. A scanner has a shallow depth of field. A card that bows up off the platen is both blurred and slightly magnified along the lifted edge.
  4. A standard card is 63.5 × 88.9 mm (2.5 × 3.5 in). That is the sanity check: at 600 dpi the card should measure 1,500 × 2,100 pixels in the scan. If it does not, your dpi setting is not what you think it is.

With that, the table:

Scan resolutionmm per pixelCard width in pixels (63.5 mm)A 2.5 mm border, in pixelsA 2.0 mm border, in pixels
300 dpi25.4 ÷ 300 = 0.0847 mm750 px29.5 px23.6 px
600 dpi25.4 ÷ 600 = 0.0423 mm1,500 px59.1 px47.2 px
1200 dpi25.4 ÷ 1200 = 0.0212 mm3,000 px118.1 px94.5 px
2400 dpi25.4 ÷ 2400 = 0.0106 mm6,000 px236.2 px189.0 px

A worked measurement at 600 dpi. You count the left border at 66 px and the right border at 54 px:

  • Left % = 66 ÷ (66 + 54) = 66 ÷ 120 = 55.0%, so the card is 55/45 left to right.
  • In millimeters: 66 × 0.0423 = 2.79 mm and 54 × 0.0423 = 2.29 mm. Combined 5.08 mm, difference 0.50 mm, physical shift 0.25 mm.

Notice you never needed the millimeters. The ratio is a pixel count divided by a pixel count and the units cancel, so you can measure entirely in pixels and only convert at the end if you want a number to compare against a caliper reading. That is also why the dpi setting does not have to be exact for the ratio to be right, only consistent within the one image.

What resolution actually buys you

The useful question is not "how fine is a pixel" but "how far does the answer move if I misplace an edge by one pixel". The worst realistic case is misplacing both sides by one pixel in opposite directions, which is what happens when the detected card outline is shifted slightly. Starting from a truly 50/50 card:

Resolution2.5 mm borders (5.0 mm combined)Reading after ±1 px on each side2.0 mm borders (4.0 mm combined)
300 dpi29.5 px each30.5 ÷ 59.0 = 51.7%, ±1.7 points±2.1 points
600 dpi59.1 px each60.1 ÷ 118.2 = 50.9%, ±0.9 points±1.1 points
1200 dpi118.1 px each119.1 ÷ 236.2 = 50.4%, ±0.4 points±0.5 points
2400 dpi236.2 px each237.2 ÷ 472.4 = 50.2%, ±0.2 points±0.3 points

Read that against the grade lines. PSA's published 10 and 9 front lines sit 5 points apart; TAG's Pristine line of approximately 51/49 sits about 4 points inside its Gem Mint line of approximately 55/45. A 300 dpi scan wobbles by nearly two points before you have made any other mistake, which is a third of a grade band. A 600 dpi scan wobbles by under one point. 1200 dpi is the point of diminishing returns: beyond it your pixel-picking and the softness of the printed border edge dominate, not the sampling.

That last point deserves emphasis. Under enough magnification the outer edge of a printed border is not a mathematical line; it is a transition a few pixels wide where ink coverage falls off. At 2400 dpi you can see that transition, which does not make you more accurate — it makes you responsible for picking a rule and applying it identically on all four sides. Consistency beats magnification.

The phone camera side of the arithmetic

A phone has no dpi setting, so you derive an effective one from how much of the frame the card fills. The card is 88.9 mm tall, which is exactly 3.5 inches:

Effective dpi = (card height in pixels) ÷ 3.5

Card height in the imageEffective dpimm per pixel
4,032 px (card fills the long side of a 12 MP frame)1,152 dpi0.022 mm
2,800 px800 dpi0.032 mm
2,000 px571 dpi0.044 mm
1,300 px (card fills about a third of the frame)371 dpi0.068 mm

So on resolution alone the phone wins as long as the card is large in the frame. Digital zoom does not help: it crops and resamples, so the effective dpi is set by how close the lens actually was.

The error a scanner does not have: perspective

This is the whole ballgame. If the phone is not parallel to the card, the near edge is magnified relative to the far edge. Using a pinhole model on a card tilted about its horizontal centerline:

  • At a 10° tilt, the near and far edges of an 88.9 mm card differ in distance by 88.9 × sin(10°) ≈ 15.4 mm, so each is about 7.7 mm nearer or farther than the card center.
  • With the phone 15 cm above, the near edge sits at about 142 mm and the far edge at about 158 mm, so the near border is magnified by roughly 158 ÷ 142 ≈ 1.11.
  • A card with truly equal 2.0 mm top and bottom borders now images at about 2.22 : 2.00, which computes to 52.6/47.4. A genuinely 55/45 card with the wide border on the far side reads about 52.4/47.6 and looks like a comfortable 10.
  • At 5° from 15 cm, or 10° from 30 cm, the effect roughly halves to about 1.3 points.

A 5° tilt is not visible to the eye while you are holding the phone. A scanner's geometry makes the error structurally impossible: the sensor and the glass are parallel by construction and the lighting travels with the sensor, so both ends of the card are imaged identically. That is the single reason to prefer a scan, and it is worth more than every resolution number above.

Software can partly undo perspective by detecting the card outline and re-projecting it to a true rectangle, but the correction inherits the accuracy of that detection, and on a card with a shadow along one edge or a sleeve still on it, detection is exactly what fails. The free card centering calculator lets you place the guide lines yourself for this reason; the app measures from a photo and maps the result to the PSA, BGS, CGC, SGC and TAG standards. Neither can recover information the image never contained. The full accuracy breakdown by tool is in do card centering apps and tools actually work.

Scanner problems nobody warns you about

A scan is not automatically correct. Five things go wrong:

Newton rings. Where the glossy face of a card sits in near contact with the scanner glass, you get concentric colored interference fringes. They do not move the card edge, so they rarely change a centering number, but on a full-art card whose reference is a one-millimeter frame line, a fringe sitting on that line makes it unreadable. Scanning the card face up with the lid closed, or lifting it a hair off the glass, breaks the contact.

Lid pressure. Closing the lid flattens a curled card, which is good for the measurement and bad for a card that already has a bend. On a warped holo, press only enough to make it lie flat and stop. Never weight the lid.

Edge glow. Scanner lamps are bright and the light bounces off the white underside of the lid, so a white-bordered card can lose its edge entirely, and even where it survives the bright background bleeds a pixel or two into the card and pushes your detected edge inward. Put a matte dark sheet behind the card so the cut edge is unambiguous — a light one for a black-bordered card.

Dust, glass marks and interpolated resolution. At 1200 dpi a dust mote is a visible blob and a fingerprint under the glass is a soft grey band; neither matters until one lands on a border edge, so wipe the platen first. And setting 4800 dpi on hardware that resolves less only makes a bigger file — use the optical figure and check by counting the card.

One thing a scanner is bad at: it tells you nothing about gloss, holo pattern, print lines or surface scratches, because it lights the card from one fixed geometry. Not a centering problem, but a submission check still needs a raking-light look with your eyes.

Phone problems you can actually fix

  • Tilt. The dominant error, quantified above. Use the camera's level or grid, brace your elbows, or lay the phone on a small stack of books with the lens over the card.
  • Lens distortion. Wide-angle lenses bend straight lines near the frame edges. Keep the card comfortably inside the frame, and prefer the main (1×) lens over the ultra-wide.
  • Shadows and glare. A card lit from one side throws a thin shadow along the far edge that reads as extra card, and a hotspot on foil can erase a border line locally. Light from two sides at about 45°, or use diffuse daylight. No flash.
  • Curl. A bowed card reproduces the tilt problem in miniature along one edge. Flatten it, or scan it.
  • Sleeves. A penny sleeve edge sits outside the card edge and is frequently the crispest line in the photo, so automatic detection locks onto it. Take the card out. What each holder does to a measurement is covered in measuring centering through sleeves, toploaders and slabs, and the full shooting protocol is in how to photograph a card so its centering measures correctly.

Protocols

Flatbed scan, front and back:

  1. Wipe the glass. Take the card out of any sleeve and handle it by the edges.
  2. Place a matte sheet behind the card that contrasts with its border color.
  3. Set 600 dpi minimum, 1200 dpi preferred, no auto-crop, no sharpening, no auto color correction. Save as PNG or TIFF; JPEG compression blurs exactly the high-contrast edge you are measuring.
  4. Close the lid with normal weight only, then verify the card measures 1,500 × 2,100 px at 600 dpi (or 3,000 × 4,200 at 1200).
  5. Count pixels edge-to-border at two points per side, roughly 15 mm from each corner, and average. Two points catch a tilted or diamond-cut print.
  6. Repeat for the back. PSA lists 75/25 for a 10 and BGS 55/45 for a Pristine, so the back can decide the grade on its own.

Phone photo: card out of the sleeve and flat on a matte contrasting surface; diffuse light from two sides and no flash; main lens directly over the card center, phone parallel to the table, card filling most of the frame but not touching the corners; tap to focus on a card edge and shoot at full resolution with no digital zoom. Then take three shots and measure all three — agreement within a point means the geometry was good, disagreement means the photo is the problem rather than the card. Same two-points-per-side rule, and do the back.

Which one to use

SituationUseWhy
Sorting a box of raw cards into keep and passPhoneSpeed. You only need to know whether centering is roughly fine.
Deciding whether one card is worth a grading feeScanner at 1200 dpi, or a careful phone photo measured three timesA point of accuracy decides a grade line that decides the fee.
A card within two points of a published line, or a thin-reference full-artScannerYou need the error smaller than the question; a 1 mm frame line is 24 px at 600 dpi and 47 px at 1200.
Documenting a card for a listing or a disputeScannerA flat, square, repeatable image other people can measure themselves.
Checking gloss, print lines or surface under raking lightPhoneA scan flattens exactly the information you are looking for.

If you own a flatbed, there is no reason to measure a submission candidate any other way. If you do not, do not buy one just for this: a careful straight-down photo with the card filling the frame lands in the same accuracy class. The step-by-step for every method, including calipers and rulers, is in how to measure card centering at home.

Bottom line

  • The scanner's advantage is geometry, not resolution: fixed parallel sensor, no perspective error, and a scale you know in advance.
  • mm per pixel = 25.4 ÷ dpi. At 600 dpi that is 0.0423 mm, at 1200 dpi 0.0212 mm. Assume optical dpi, an unresampled file viewed at 100%, and a card lying flat.
  • Scan at 600 dpi minimum, 1200 dpi ideally; a one-pixel error per side costs about 0.9 points at 600 dpi and 1.7 points at 300 dpi.
  • A phone matches a 1200 dpi scan when the card fills the frame (about 1,150 effective dpi), but a 10° tilt at 15 cm shifts a centered card by roughly 2.6 points and a 5° tilt by half that.
  • Scanner-specific traps: Newton rings on gloss, lid pressure on curled cards, edge glow against a white lid, dust at high dpi, and interpolated resolution that adds nothing.
  • Ratios are pixel counts divided by pixel counts, so you can work entirely in pixels; measure two points per side, do the back, and verify 1,500 × 2,100 px at 600 dpi as a sanity check.

Frequently asked questions

Is a scanner more accurate than a phone for measuring card centering?

For geometry, yes, and the reason is not resolution. A flatbed scan is taken at a fixed distance with the sensor parallel to the glass, so there is no perspective error at all, and the scale is known in advance from the dpi setting. A phone can match or beat a scanner on raw resolution but has to be held parallel to the card, and a tilt of only a few degrees moves the measured ratio by a point or two. If you own a flatbed, scan. If you do not, a straight-down photo with the card filling the frame is close behind.

What dpi should I scan a card at to measure centering?

600 dpi is the practical minimum and 1200 dpi is comfortable. At 600 dpi one pixel is 25.4 divided by 600, which is about 0.042 mm, so a 2.5 mm border is roughly 59 pixels and being off by one pixel on each side moves the ratio by under a point. At 300 dpi one pixel is about 0.085 mm and the same one-pixel error is worth about 1.7 points, which is enough to blur the line between a 55/45 card and a 57/43 card. Going above 1200 dpi mostly buys file size.

How do I convert scanned pixels to millimeters?

Millimeters per pixel equals 25.4 divided by the scan dpi, because one inch is exactly 25.4 mm. At 600 dpi that is 0.0423 mm per pixel, at 1200 dpi 0.0212 mm, at 2400 dpi 0.0106 mm. Multiply your pixel count by that figure. You can also skip the conversion entirely: the centering ratio is a pixel count divided by a pixel count, so the units cancel and you can work in pixels the whole way.

Why does my scan show colored rings on a holo card?

Those are Newton rings, an interference pattern that appears where two very smooth surfaces sit in near contact, in this case the glossy card face and the scanner glass. They do not move the card edge, so they rarely change a centering measurement, but they can obscure a thin frame line on a full-art card. Scanning the card face up under the lid, or lifting it very slightly off the glass, breaks the contact that causes them.

Can I scan a card inside its sleeve or toploader?

You can scan through a penny sleeve for a quick screen, but the sleeve edge sits outside the card edge and is often the sharper of the two lines in the scan, so it is easy to measure the sleeve by mistake. A toploader adds glare and lets the card sit crooked inside the holder. For any measurement you are basing a submission on, scan the bare card on a dark matte backing sheet.

Does a phone camera have enough resolution to measure centering?

Easily, provided the card fills the frame. A card is 88.9 mm tall, which is 3.5 inches. If the card spans 2,800 pixels of the image, that is an effective 800 dpi, better than a 600 dpi scan. If the card only fills a third of the frame at about 1,300 pixels, you are down to roughly 370 dpi and digital zoom does not recover it. Resolution is the smallest of the phone's error sources; tilt is the biggest.

Sources & further reading

#scanner#phone-camera#dpi#pixel-measurement#measurement-accuracy#centering-tool#how-to-measure