Image to Excel conversion reads a photo or screenshot of a table with OCR and writes the cells as rows and columns. A sharp, straight, well-lit picture of just the table gives the best result. Always check the rows against the picture, because OCR can misread digits and shift cells.
Which image files convert to Excel?
Common picture formats such as JPG and PNG convert in almost every tool, while support for others depends on the tool. Check the tool's upload list, and when a file is refused, change its format instead of retaking the picture.
The usual formats
- PNG is what screenshots normally save as. It keeps hard edges around text, which suits OCR well.
- JPG is the usual format for phone photos and scans. It compresses the picture, and heavy compression blurs small digits.
- HEIC is the default photo format on many iPhones. Support is patchy, so if a tool rejects it, export the photo as JPG first, or set the camera to save the most compatible format.
- PDF is a different job when it holds a scan. Converting a scanned PDF to Excel covers it.
Microsoft's help page for Data From Picture lists no supported file types or size limit. If you use that tool, try a small file first, and see the page on Excel's own Data From Picture for where it exists.
Format is not the main factor
Resolution and sharpness decide the outcome far more than the extension. A sharp JPG beats a blurred PNG. When you have a choice, keep the original file and avoid re-saving it through chat apps or messaging, which often shrink and recompress pictures.
Screenshots vs phone photos for image to Excel
A screenshot gives the cleaner result when the table is already on a screen, because it is flat, square and sharp, with no shadows or perspective. A phone photo is the right choice only when the table exists on paper and no digital copy can be had.
Why screenshots convert better
A screenshot is made of crisp pixels laid out in a perfect grid. Every line is straight, every character is rendered the same way, and the lighting is uniform. OCR does best with exactly those conditions. A table captured from a web page or a PDF viewer on screen usually converts with far fewer misreads than the same table photographed from paper.
For the best screenshot:
- Zoom the page so the type is at least normal reading size before you capture it.
- Capture only the table, not the browser frame around it.
- Capture a region that fits in one screen. A long table is better captured in several overlapping pieces than in one shrunken one. Overlaps mean you delete repeated rows afterwards.
When a photo is the only option
Paper is the usual reason. A photo has to fight the page's curl, the lighting and the angle. Use the checklist below, and read scanning paper pages to Excel if you have a scanner or copier near you, because a scan is flatter and more even than a photo.
If the table is a bank statement, note that the statement may be available as a digital PDF from the bank. A digital PDF skips image reading entirely.
A photo checklist
Treat this list as the shot list for every table photo:
- Page lies on a flat surface, held down at the corners if it curls.
- Even daylight from a window, with no hard shadow across the page.
- No flash, which creates a bright patch that bleaches the print.
- The page fills the frame, with the table edge to edge.
- One page per photo, never two pages side by side.
- The camera is held parallel to the page, looking straight down.
- Text is in focus. Tap the screen on the print and zoom in to check.
Cropping an image to the table area
Yes, cropping to the table helps. OCR has fewer things to confuse with data when the picture holds only the table. Logos, letterheads, addresses and page margins can be read as stray rows, and Microsoft's own guidance for its picture import is to include only the data you want.
What to cut and what to keep
Keep the header row, every data row and the full width of the table. Cut the logo, the page border, the footer and any text above the header. Leave a thin margin of white around the table, because a crop that clips the edge of the first or last character costs you a digit.
Cropping in practice
Most phone cameras and photo apps have a crop tool. Crop, then save a copy. Do not stretch the crop to a different shape, which distorts the letters, and do not crop after the file has been compressed repeatedly. If the camera app has a document mode, it crops and flattens at the same time.
A cropped table with the headings cut off needs one more step. The converter will not know which column is the date and which is the amount, so add headings by hand if they were lost.
What should you check in image to Excel output?
Check that the shape is right, then check the numbers. A fast review of image to Excel output has four parts: confirm the columns, confirm the row count, test that the numbers are numbers, and compare a few figures with the picture.
The quick check, in order
- Row count. Count the rows in the picture and in the sheet. A missing row is a missing transaction.
- Column count and headings. Every value should sit under the right heading, with no spill into the next column.
- Numbers or text. Select the amount column. If the status bar shows a Sum, they are numbers. If it shows only a Count, some cells are text, often from a letter-for-digit misread like "l" for "1".
- Three random rows. Compare date, description and amount with the picture.
- The largest amounts. Compare them all.
- A total. If the original shows a total or a closing balance, make the sheet reproduce it.
Mistakes that cost the most time
- Photographing a screen with a phone. This adds glare and moire stripes. Take a screenshot instead.
- Converting a long table in one shrunken image. Small type is where digits go wrong. Split the table into pieces at a readable size.
- Skipping the headings. Without them you must guess which column holds debits and which holds credits.
- Pasting the result over your books before checking. Convert into a new sheet, check it there, then copy the clean rows across.
- Assuming a second attempt will behave like the first. Light changes, so a retake can read better or worse. Compare both before you pick one.
Example of a catch
Here is an invented example. The picture shows a debit of 1,250.00. The sheet shows 125000. The comma was dropped and the decimal point was missed. The number looks plausible by itself, and only a column total or a balance exposes it. Decimal and separator errors in OCR'd amounts covers this failure in detail, and the same pattern holds for any printed table.
When the table is a bank statement
A photo of one statement page can be converted for a quick look. For a statement you will keep in your records, a photo is a weak source. FastBankConvert reads digital statement PDFs and tells you before download when the balances do not add up; the product is getting ready to open, so see the get started page for access. For phone-based tools more broadly, apps that convert PDF to Excel on a phone compares the options.
Glare, shadows and curved pages in table photos
A photo converts badly when light or geometry changes the look of the print. Glare washes out digits, shadows darken part of the table into a blob, and a curved page bends rows so columns no longer line up.
Fixing each problem
| Problem | What it looks like | Fix |
|---|---|---|
| Glare from glossy paper | A white patch over some digits | Tilt the page or move the lamp, then shoot again |
| Your own shadow | A grey band across the table | Shoot from a window side, not with your body between page and light |
| Curved page near a binding | Rows bend and the left columns slant | Press the page flat, or scan a loose sheet |
| Perspective | The table looks like a trapezoid | Hold the camera parallel to the page |
| Blur | Soft edges on digits | Tap to focus, hold still, use a brighter room |
| Low light | Grainy digits | Add daylight, avoid flash |
Why the conversion fails
OCR separates ink from paper by contrast. Glare and shadow damage that contrast. Curvature breaks the straight rows and columns that the table step relies on, which builds columns by comparing word positions. When you see a pattern of errors that follows the light or the curve across the sheet, retake the picture instead of correcting cell after cell. A new photo takes thirty seconds, and hand-fixing fifty cells takes much longer.