Photo of a Computer Screen: Why the Text Reads Badly

A photo of a computer screen converts poorly because the camera photographs the monitor itself: its grid of tiny lights, its reflections and its refresh cycle, with the words mixed in. Recognition software then has to separate letters from stripes and glare. A screenshot or an exported file avoids every one of those faults.

For a client looking at a bank portal, a payroll screen or a chat window, the phone is the fastest capture within reach. The result is readable to a person and troublesome for software, and knowing why helps you decide whether to work with the picture or ask for something else.

Why does a photo of a computer screen look striped?

A photo of a computer screen looks striped because two grids are laid over each other: the monitor's rows and columns of pixels and the camera sensor's own rows and columns. Where the two fall in and out of step, the picture shows wavy bands, rings or colored ripples. The effect is called moiré.

Every flat screen builds its image from small dots, each made of red, green and blue parts. Your eye blends them at normal viewing distance. A phone camera, held close, records the dots individually, and its sensor is a grid as well. Two fine grids at slightly different spacings or angles produce a third, coarser pattern that belongs to neither.

Three things about that pattern matter when the picture holds text:

• It moves. Shift the phone a little nearer, farther or more tilted and the bands change width and direction. Two photos of the same screen can look quite different.
• It has color. Because each screen dot is three colored parts, the ripples come out as rainbow fringes, including across text that was plain black on white.
• It crosses the letters. The bands run through words, so parts of one letter are lighter and parts darker. Thin strokes can vanish where a dark band meets them.

Zooming in on the photo makes this plain. The letters are not smooth shapes. They are built from the screen's dots, with the ripple running over them.

Can OCR read text photographed off a monitor?

OCR can read text photographed off a monitor some of the time. Large, dark text on a plain light background, photographed square on, may come through with few errors. Small text, colored text, tables and anything crossed by bands or glare will have more. Treat the output as a draft to check against the picture.

What decides the result:

Figures deserve particular suspicion. A band across a row of account numbers can turn an 8 into a 3 or drop a decimal point, and nothing in the converted text marks where that happened. Schema Law's homepage says recognition is strongest on clean print, and a photographed monitor is some distance from clean print. If you do upload one, compare every number, date and name with the image.

The photo also has less to offer than it seems. A picture of a monitor shows only what was on screen at that moment. Anything scrolled out of view is absent, and a reader of the converted text cannot tell.

If the window shows a check a client may need as documentary evidence, shoot the whole window and then anything scrolled away. Keep every photograph, because extracted text misses whatever was off screen. Matching file names help authentication, and the capture time starts the chain of custody.

Taking a usable photo of a screen when nothing else is allowed

If a photo is the only capture open to you and you are permitted to take it, enlarge the text on screen, hold the phone parallel to the monitor, step back until the stripes fade, and take several frames. Then check each one at full zoom before you leave the screen.

First, the permission. A public terminal at a clerk's office, a records kiosk or an employer's system may forbid photographs or copies of what it shows. Ask, and follow the answer. If the system blocks screenshots on purpose, a phone photo is not the fix; a request to whoever controls the records is.

Where a photo is allowed:

1. Make the text bigger on the screen. Use the program's zoom so the words fill the window. Bigger letters survive the ripple better than small ones.
2. Use a light background if you can choose. White letters on a dark theme can photograph with a glow around each stroke.
3. Clean the glass. A dry, soft cloth removes the smears a camera picks up.
4. Kill the reflections. Turn off the lamp behind you or tilt the monitor until windows and lights leave the glass. Flash off.
5. Square up. Center the lens on the middle of the window with the phone's back parallel to the screen.
6. Change distance to tame the stripes. Moiré depends on how the two grids line up. Move the phone a little closer or farther and watch the bands on the phone's display. Take the photo where they are weakest.
7. Take three frames. Slight differences in position move the bands, so a word spoiled in one frame may be clean in another.
8. Capture the context. Include the address bar, the account name, the date shown on the page and the scroll position, in a separate frame if needed.
9. Scroll and repeat with overlap. For a long page, keep two or three lines from the previous screen at the top of the next.

Do not edit the pictures to make them prettier. Send them as taken, at full size.

What should you ask for instead of a picture of a screen?

Ask for the thing behind the screen: a downloaded statement, an export or a page saved as a PDF. If none exists, ask for a screenshot taken on the device itself. A screenshot copies the exact pixels the device drew, with no camera, no stripes and no glare.

In order of preference:

1. The original file. A portal's download button, a message export, a saved report. This can contain real text and needs no recognition at all.
2. Print to PDF. The browser's or program's print command with a PDF as the destination captures the whole page, not just the visible part.
3. A screenshot. Each platform documents its own shortcut: - On an iPhone with Face ID, Apple's instructions for taking a screenshot on an iPhone say to press the side button and the volume up button at the same time. - On Android, Google's help page on taking a screenshot on an Android device says to press the Power and Volume down buttons at the same time. - On a Mac, Apple's page on taking a screenshot on a Mac gives Shift, Command and 3 for the entire screen. - On Windows, Microsoft's page on using Snipping Tool to capture screenshots gives the Windows logo key with Shift and S.

A long page on a phone can be captured in one piece, and the post on saving a full-page screenshot as a PDF has the steps. Once you hold a real screenshot, the section on getting text out of a screenshot of a document explains how to convert it.

A short request you can paste into an email to a client: "Please do not photograph the screen. Download the statement as a PDF if the site offers it. If it does not, take a screenshot on the same device and send it without cropping."

Glare, fingerprints and refresh bands on a photographed monitor

Besides moiré, a photographed monitor can show reflections of the room, smears on the glass and dark horizontal bars from the screen's refresh. Each one hides or distorts text in a different place, and each has a simple check before you rely on the picture.

Refresh bands come from timing. A screen redraws many times a second and may also pulse its backlight, while a phone camera reads its sensor a line at a time. When the two fall out of step, part of the frame catches the screen in a dimmer moment. Taking several frames is the practical answer, since the bar lands in a different place each time.

Glare and reflections follow the rule for any glossy surface: move the light, the screen or yourself until the bright spot leaves the text. Words on other shiny objects in a scene raise the same problems, and the post on text in a photograph of a sign or label covers those.

Whatever you end up with, keep the photo beside any text made from it. If the picture is the only record of what the screen showed, the image is the record and the text is a finding aid. Schema Law accepts PNG and JPEG uploads as well as PDF, and the how it works section shows the steps from upload to export.