OCR Faded Text: Thermal Paper, Weak Toner and Pencil

OCR can read faded text when the scan still shows the print as continuous strokes, however pale they are. It fails when a black and white scan has already thrown the faint marks away. Scan the page in grayscale or color, convert that file, proofread the pale passages, and copy any original that is still fading.

Can OCR read faded text on a scanned page?

Yes, if the pale print is still in the image as unbroken gray strokes. Recognition needs a difference between ink and paper, and it copes with a small difference better than with none. What it cannot do is read marks that the scan did not record, or rebuild letters whose strokes have gaps.

Three questions sort a faded page before you convert it.

Is the fading on the paper or only in the scan? Put the paper beside the screen. If the page in your hand is darker and more complete than its picture, the scanner settings lost the print and a second scan will get it back. If the paper is as pale as the scan, you are working at the limit of what exists.

Is the fading even? A page that is pale all over can be handled with one setting. A page that is dark on the left and ghostly on the right, or pale along an old fold, has no single setting that suits both halves.

Is the page still changing? Toner on bond paper stays put. Some thermal prints and some inks do not. If the original is on its way out, capture comes before everything else.

Schema Law's homepage answers the question "How bad can the scan be?" with "Upload it as it is and look at the result", and says recognition is strongest on clean print and weaker on faint faxes. It adds that nothing marks a passage as uncertain. So a pale page gives you text that looks finished whether or not it is, and the check against the original is yours to do. The questions section of the homepage has the full wording.

Why does faint print vanish in a black and white scan?

A black and white scan keeps one fact about each dot on the page: darker than a cutoff, or lighter. Faded print is lighter than the cutoff, so the scanner files it with the paper and writes white. The letters are not dimmed in the file. They are absent from it, and no later step can bring them back.

Put rough numbers on it. Say the scanner measures brightness from 0 for black to 255 for white:

• Clean paper measures about 235.
• Fresh laser toner measures about 30.
• A faded thermal print measures about 170.
• The scanner's cutoff sits at 128.

Fresh toner is far below the cutoff and comes out black. The faded print, at 170, is on the paper's side of the line. It was plainly visible to the person feeding the scanner, sixty-five steps darker than the sheet, and it is gone from the file.

Partial loss is worse than total loss

A page that drops out completely is at least obvious. The harder case is print that sits near the cutoff, where the thick parts of each letter fall just under it and the thin parts just over.

• An "m" keeps its three stems and loses the arches, and reads as "iii" or "ill".
• An "e" loses its crossbar and reads as "c".
• A decimal point or a comma, which is the smallest mark on the line, is the first to go.
• Whole lines vanish from the paler half of an unevenly faded sheet while the darker half converts normally.

The output then looks like a complete page with a few odd words, and a reader has no reason to suspect that a third of the content is missing.

Recognizing the damage in a file you were sent

Zoom to 400% on a broken letter. If every dot is pure black or pure white, with no grays at the edges, the file is bitonal and the cutoff has already been applied. A darker or lighter view in your PDF reader changes nothing, because there are only two values to work with. A page that arrived by fax is in this state, because a fax is sent as black and white dots.

When the paper is available, the cure is a new scan that keeps the shades. The guide to grayscale versus black and white scanning explains how to choose between grayscale and color and what each costs in file size.

Keeping the shades also leaves a better decision to the software. OCR engines make their own ink-or-paper choice internally, and some can make it region by region. The Tesseract documentation on improving the quality of OCR output notes that "Tesseract 5.0.0 added two new Leptonica based binarization methods: Adaptive Otsu and Sauvola." An engine can only use a method like that if the grays are still in the file it is given. A bitonal scan has made the choice already, once, for the whole sheet.

The same loss happens to the back copies of multi-part forms, where the mark was light to begin with. The post on OCR for carbon copies and carbonless forms covers those.

Thermal paper receipts and fax rolls that have gone pale

Thermal paper has no ink or toner on it. The printer heats a coated sheet and the coating darkens where it was heated. The image can fade, or the whole sheet can darken, long after printing. Till receipts, card slips and old fax rolls are the thermal prints a legal file collects.

The United States National Archives described the problem for federal agencies in NARA Bulletin 96-03 on preserving facsimile transmissions. The bulletin dates from 1996 and expired in 1998. The Archives marks the page as superseded, warns that its information is no longer accurate as guidance, and keeps it online "only as a technical and historical reference". It is cited here as a record of what the Archives observed about thermal paper, not as current guidance. Its paragraph 3 says "The images on thermal paper are extremely unstable and may begin to deteriorate in as few as six months." It goes on to say that thermal papers "are sensitive to heat and light, will react with chemicals found in ordinary office environments (including markers, cosmetics, and some types of plastic folders), and may contain impermanent dyes." The stated result: "the text is likely to fade and become illegible or the whole paper surface may darken, making the image indistinguishable."

The bulletin is decades old and was written about fax paper. How long a given receipt lasts depends on the paper and where it was kept. Treat the bulletin as a description of what can happen, and look at the receipt in front of you.

Two ways a thermal print fails

The bulletin's sentence names both. In the first, the text fades toward the color of the paper and you get pale gray on white. In the second, the sheet darkens and you get gray on gray or black on dark brown. Grayscale handles the first. The second is a color problem: text and background may differ more in hue than in brightness, so a color scan gives software something to separate.

Handling a thermal print

• Do not highlight it. The bulletin lists markers among the things thermal paper reacts with.
• Keep it away from heat. That includes a laminator, a sunny windowsill and the top of a warm machine.
• Do not tape over the print or store it in a soft plastic sleeve without checking first, given the bulletin's warning about plastic folders.
• Scan it flat on the glass. A receipt is narrow and curls, and a feeder can skew or crease it.
• Scan it in color at a higher resolution than you use for letters, because receipt type is small.

What to pull from a faded receipt

Suppose a client in a fictional reimbursement dispute, Ines Varga, hands over forty till receipts from 2022 in an envelope. Half are pale. Convert them all, then check by hand the five fields that carry the claim:

1. Merchant name.
2. Date and time.
3. Each line amount she is claiming.
4. The total, and the tax line if it matters.
5. The last four digits of the card.

Where a field cannot be read on the scan or the paper, write "[illegible]" in your schedule and say so. A bank or card statement for the same date can supply the amount and the merchant, and then the receipt and the statement support each other.

Low-toner printouts and pencil entries with streaks and gaps

Some print was weak on the day it was made. A laser printer short of toner leaves pale vertical bands. An inkjet with a blocked nozzle leaves thin horizontal gaps through every line. Pencil is gray, thin and shiny. None of these has faded, but all of them sit near the scanner's cutoff in the same way faded print does.

Low toner and inkjet gaps

A toner streak runs down the page in the direction the paper travelled, at the same distance from the edge on every sheet of the print run. Whatever sits in that band is weakened on every page. If the band crosses a column of figures in a forty-page statement, every figure in that column is at risk, and the errors will be of the same kind each time.

Inkjet banding runs across the page. A missing row of dots through the middle of a line turns "8" into something like "0" over "0", and "B" into "E"-like fragments. The letters are tall enough, but a slice is missing from each.

For both, use the pattern. Mark the band on the first page, and proofread that band on all pages before anything else. Impact printers have their own version of this, and the post on dot matrix print covers draft mode, near letter quality and worn ribbons.

Pencil

Pencil turns up where someone filled in or corrected a printed page: figures entered on a worksheet, tick marks on an inventory, a corrected date in a margin, initials beside a change. Two things make it hard. Graphite is lighter than ink, so a black and white scan can drop it. And a pencil entry is handwriting. Schema Law's homepage says recognition is weaker on handwriting, so plan to read pencil entries yourself even from a good scan.

The specific trap is the printed form with pencil answers. The printed labels convert cleanly. The pencil answers vanish or come out as fragments. The result reads like a blank form, and someone searching the file will conclude that the form was never filled in.

A check for pages with weak marks

• Scan in grayscale so that pencil and pale print survive.
• Count the entries on the page image and count them in the text. A form with twelve filled boxes should not produce three values.
• Read every pencilled figure against the image and type it in yourself.
• Note in the working copy which entries were typed by a person from the image.

Should you darken a faded scan before running OCR?

Only when the scan holds grays to work with, and only on a copy. Darkening widens the gap between pale ink and paper, which helps when the print is faint but whole and the paper is clean. It does nothing for a bitonal file, and on stained or yellowed paper it darkens the background along with the print.

Convert the page unedited first. If the pale passages come out acceptably, there is nothing to fix and no altered image to account for.

When darkening helps and when it hurts

How far to go

Watch the paper, not the print. Raise the darkness or move the mid-tone control in small steps, and stop at the first step where the blank areas of the page begin to show grain or specks. Past that point you are adding marks, and specks near a figure get read as punctuation.

Then test. Convert the adjusted copy and compare a fixed passage, including its figures, with the result from the unedited scan. Keep the version with fewer errors.

Rescanning darker is better than editing darker

If you hold the paper, change the capture and leave the editing alone. A scanner's brightness or exposure setting changes what the sensor records. An image editor can only rearrange what was recorded. A second scan in grayscale, a little darker than the default, holds real detail that no slider can produce from the first file.

Any adjusted image is a working copy. Keep the first scan, or the file you received, unchanged, and say in your note that the text was converted from an image whose contrast was increased.

Copying a fading original before it gets worse

Scan a fading page on the day it reaches you. An original that is still changing will never again be as legible as it is now, so the first good scan becomes your best record of what the page said. Keep the paper too, and write down when the scan was made and what state the original was in.

The National Archives bulletin gave federal agencies that instruction for fax paper in paragraph 5: "All thermal paper facsimiles that are Federal records should be copied on plain paper at the time of receipt." That direction was addressed to agencies and is no longer in force, since the Archives marks the bulletin as superseded, but the reasoning carries over to any file: copy first, because waiting only costs legibility.

Do the same the day a faded will arrives. You record the decedent's name, the date, and any legacy you can still see, and you note that this intake scan is the one you will use if estate administration or a probate filing later asks what the page said when you received it.

An intake routine for fading originals

1. Scan before handling further. Color, flat on the glass, before anyone staples, highlights or sorts the pages.
2. Open the scan beside the paper. Confirm that everything legible on the paper is legible on screen. If not, rescan darker now.
3. Name the file with the capture date. "Varga receipts scanned 2026-03-14.pdf" settles later arguments about when the image was made.
4. Write a condition note. For example: "Thermal receipts, 40 items. Items 7, 12 and 31 faint at intake; totals legible on paper and on scan."
5. Store the paper cool, dark and flat. A paper folder in a drawer, away from windows and machines.
6. Convert from the intake scan. Later scans of the same paper will be of a paler page.

Keep the paper

A scan does not make the original disposable. Whether a copy can stand in for an original, and what you must retain and for how long, depend on the evidence rules of your court and on your own retention obligations. Check those before anything is thrown away. If the original keeps fading in the file, your dated scan and condition note show what it said when it arrived.

Ask clients early

A receipt can sit in a client's glovebox for months before a dispute brings it to you. If your intake letter asks for receipts, ask at the same time that they be sent or scanned now and kept out of cars, wallets and plastic sleeves in the meantime. Once the scans arrive, the pricing page shows the page limits that apply to one document on each Schema Law plan, which matters when forty receipts are combined into one file. A soft scan of a pale receipt is two problems at once, and the post on OCR for a blurry scan covers the second.