How to Read a Faded or Blurry Scanned Bank Statement

A client drops a statement in your inbox and half the numbers look like they were printed underwater. You can’t tell if that amount is a 3, an 8 or a 9. Most of the rows are fine. A few are barely hanging on. One corner has a branch stamp parked right on top of a balance.

You still need the whole thing in Excel by tomorrow. Sitting there typing every figure by hand is going to eat your evening.

Old printouts, photocopies of photocopies, and phone photos all do this. Getting the text out is only half the battle. You still have to make sure the numbers in your spreadsheet actually match what was on the original page.

Why these scans are such a pain

Open the PDF and try highlighting an amount. If the text lights up, the file already has real text and extraction is usually pretty straightforward. If the page behaves like a photo, it’s just a scanned image. Your computer sees pixels, not numbers.

OCR tries to turn those pixels into characters. It does okay on clean, sharp text. It struggles with faded ink, smudges, or rows that look like they slid sideways. The usual suspects:

- Faded ink. The thin digits disappear first, and the decimal points often go with them. Lose a decimal and Rs. 1,250.00 suddenly becomes Rs. 125,000.

- Dot-matrix printing. Those characters are made of little dots. After a couple of photocopies the dots start vanishing or merging, and a 6 starts looking a lot like a 5.

- Photocopies of photocopies. Each generation adds more noise and loses more detail.

- Phone photos. Shadows, glare, curled paper, and shooting at an angle make the rows slope and leave one side of the page darker than the other.

- Stamps and handwriting. A stamp right across a figure can hide the exact number you need. Image enhancement tools can help a bit, but they can also wipe out the faint marks you’re trying to save. Always keep the untouched original.

The sneaky part is that a messy extraction can still look tidy. An amount can land one column over or next to the wrong description, and a quick glance at the finished sheet won’t catch it.

First thing: just ask for the e-statement

Before you start wrestling with a bad scan, ask the client if they can send the original electronic statement. Most Sri Lankan banks (Commercial Bank, Seylan and others) offer e-statements usually password-protected PDFs emailed to registered customers. An electronic PDF is almost always cleaner and easier to work with than a scan of a printed page, whether it’s a current account or a savings account.

If they only have paper, you’re stuck with whatever quality the scan gives you.

If you have to scan it yourself

A few quick checks make a real difference.

Resolution: start at 300 DPI. That’s the usual office standard for OCR. For faded or low-contrast pages, greyscale at 300–400 DPI works better (400 is often recommended). The right setting depends on the document, so don’t be afraid to try a couple of versions.

Colour mode: greyscale is usually friendlier to faint ink. Use colour if stamps or coloured marks matter.

Alignment. Press the page flat so nothing curls, keep all four edges visible, and double-check that no digit gets chopped off at the margin. If you’re photographing it with your phone, shoot straight down in even light, then zoom in to make sure the decimal points are still clear.

Compression. Heavy JPEG compression blurs small digits together. Don’t keep resizing and re-saving the image.

Enhancement tools.  Programs like Adobe Acrobat can straighten pages and clean up background noise. Sometimes they help a lot. Sometimes they erase the faint marks you actually needed. Always keep an unedited copy and compare the two.

Can OCR or AI actually turn a blurry statement into Excel?

Yes , but how well it works depends on the scan quality and the tool.

OCR is good at spotting characters. Better document tools also try to sort dates, descriptions and amounts into the right places. The classic mistake is when the software reads everything correctly but still drops the amount onto the wrong row or into the wrong column.

General AI tools have their own quirks with financial documents. They sometimes quietly push a hard-to-read character toward whatever looks most plausible, and numbers are especially vulnerable to this. An AI-generated table can look perfectly clean even when some of the figures are wrong. Unless the software flags the problem, you might not notice until later.

Whatever tool you use, always check the extracted figures against the original before you start using the spreadsheet for real work.

The simple way to check the numbers: use the running balance

If the statement shows a running balance after each transaction, you’ve already got a built-in test. Each new balance should equal:

previous balance − debit (or + credit) = new balance

Quick example (all in rupees):

- Opening balance: Rs. 48,250.00  

- Debit: Rs. 3,150.00  

- Balance printed on the statement: Rs. 45,100.00  

48,250 − 3,150 = 45,100. It matches.

Now suppose the software reads the debit as Rs. 3,750 instead. The calculated balance becomes 44,500 — Rs. 600 short of the printed figure. You immediately know which row to re-check on the original page instead of reading the whole statement again.

This row-by-row check is useful, but it has limits:

- A mismatch only tells you where to look. It doesn’t tell you the correct figure only the original page does.

- A matching balance is strong evidence, not absolute proof. Two errors can cancel each other out.

- Some statements don’t print a balance on every row.

- It works best when you treat the whole statement as one continuous document rather than separate files.

If there’s no running balance, fall back on the opening balance, the closing balance, and any printed totals of debits and credits.

Tools built for bank statements

There are converters built specifically for bank statements , including ones that work well with Sri Lankan banks. You upload the file, the tool pulls out the transactions, and you get to review the rows before exporting. Some of them even check the extracted amounts against the printed running balances and flag the ones that don’t add up. That means you only need to re-check the suspicious rows instead of going through every single figure by hand.

Once you’re happy with the results, you can export to Excel or CSV. A lot of these tools also let you export straight into Xero, QuickBooks and similar accounting software. They can handle statements from multiple banks or covering several months, and many will do basic auto-categorisation as well.

Just remember that categories are only a starting point. If you’re getting the data ready for RAMIS tax filing or SVAT tracking, take a moment to review the categories yourself against the current rules before you rely on them.

Even the best tools have limits. A badly damaged statement can still contain figures that neither you nor the software can read properly. If a digit is unreadable to you, treat it as unreadable. Always double-check unclear figures against the original, and keep the original statement next to your spreadsheet so you can quickly trace any disputed row back to the source.

 Quick checklist before you trust the data

1. Make sure the first and last pages are there and nothing’s been cut off.  

2. Compare the opening and closing balances with the original by eye.  

3. Pay extra attention to rows with stamps, folds, faded patches or unclear decimal points.  

4. Check that each description is sitting next to the right amount.  

5. Investigate any row where the balance doesn’t follow , don’t change a figure just to force the numbers to agree.  

6. Compare the total debits and credits with any summary printed on the statement.  

7. Keep the original file together with the finished spreadsheet.

These checks become even more important when you’re dealing with statements from several clients or different banks. A clean electronic PDF is almost always easier to work with than an old photocopy that needs extra scrutiny.Try it on a poor-quality statement

Got a faded or heavily photocopied statement sitting on your desk? Upload it to a converter and see how it handles the tough pages. Review the extracted transactions and check whether it flags the rows that need another look. Plenty of these tools offer a free allowance (often without needing a credit card), so you can test them out without committing to anything.