Convert JPEG to CSV

A spreadsheet, database, or data-analysis system may require values from a scanned table, receipt, or form in CSV format. Because a JPEG stores an image rather than rows and columns, useful conversion normally requires OCR and table recognition; extracting raw pixel values is a different task.

What the JPEG format is

JPEG, also called JPG, is a lossy raster-image format used by digital cameras, smartphones, scanners, screenshots, and applications such as Adobe Photoshop and GIMP. It stores colored pixels representing a visual document, receipt, form, or table; it does not normally store editable characters, cell boundaries, or spreadsheet data.

JPEG compression can discard fine details, especially after repeated saves or at low quality. A normal file converter cannot reconstruct reliable rows and values from those pixels. OCR must recognize characters, and table-recognition software must infer columns, rows, and cell boundaries.

What the CSV format is

CSV is a plain-text format in which records usually occupy separate lines and fields are separated by commas. It is supported by spreadsheet programs, databases, statistical packages, scripts, and import systems because it is small, portable, and readable without a proprietary application.

CSV does not preserve fonts, colors, merged cells, formulas, images, or page layout. A field containing a comma, quotation mark, or line break must be quoted according to CSV rules. Some receiving systems expect semicolons or tab characters instead of commas, so the required delimiter and encoding must be confirmed before export.

Convert a JPEG containing text or a table

Microsoft Excel

1. In an Excel version that supports the feature, open Data → From Picture → Picture From File and select the JPEG. On mobile, use Insert Data from Picture or the equivalent camera option.
2. Review the recognized cells in the results panel. Correct misread characters, decimal separators, dates, column boundaries, and blank cells before inserting the result into a worksheet.
3. Choose File → Save As and select CSV UTF-8 (Comma delimited) (*.csv). If the destination requires another delimiter, export or convert the file using that system's specified setting.

Excel's image import is convenient for simple, clearly aligned tables. Availability and menu names vary by Excel edition, operating system, and subscription.

ABBYY FineReader PDF

1. Open the JPEG in ABBYY FineReader PDF and select every language used in the image.
2. Run OCR, then verify the detected table areas and column boundaries. Correct values containing zeros, decimal points, currency symbols, dates, and negative signs.
3. Export to .xlsx through Export → Microsoft Excel when further correction is needed.
4. Open the workbook in Excel or LibreOffice Calc and use File → Save As to create the required CSV file.

FineReader generally provides more control for multi-column pages and complex tables than basic image-import features, but its recognition still requires manual checking.

Free local OCR

For text-oriented images, install Tesseract OCR and run:

tesseract input.jpg output --psm 6

This creates output.txt, not a structured CSV. Clean the text in a spreadsheet editor or script, identify the field separators, quote fields containing commas, and write the result as valid CSV. Tesseract alone does not reliably preserve table geometry. Camelot and Tabula are primarily PDF table-extraction tools, so they are not direct solutions for a JPEG unless the image is first placed into a suitable document and the resulting structure is verified.

Online OCR services

Google Drive can OCR an uploaded JPEG by opening it with Google Docs; copy the recognized table into Google Sheets, correct it, and use File → Download → Comma-separated values (.csv). This workflow is suitable for non-confidential images, but Google Docs may lose table structure.

Adobe Acrobat's online OCR workflow can turn an image into an editable PDF or text-based document, after which table data can be corrected and exported through a spreadsheet application. Services marketed as direct JPEG-to-CSV converters vary widely: some return plain OCR text rather than columns and rows. Check the output, retention policy, deletion controls, and supported languages before uploading an image. Do not upload identification documents, financial records, medical information, or confidential business data unless the service is approved for that data.

Convert JPEG pixels to CSV instead

If the required output is an array of image pixels rather than recognized words, do not use OCR. In MATLAB, read the image and write its numeric values with:

I = imread('input.jpg');
writematrix(I, 'pixels.csv');

A color JPEG normally produces a height-by-width-by-3 RGB array, while a grayscale image produces a two-dimensional array. Exporting a large image this way creates a very large CSV and records JPEG-decoded, already compressed pixel values; it does not recover the original camera sensor data.

Quality and compatibility checks

• OCR accuracy depends on resolution, focus, contrast, skew, handwriting, font quality, background noise, and JPEG artifacts. A clean scan around 300 dpi usually performs better than a small, compressed camera image.
• Crop irrelevant backgrounds, straighten the page, and increase contrast or use grayscale before recognition. Keep the original image so preprocessing can be repeated.
• Inspect every exported value. OCR commonly confuses 0 with O, 1 with I, decimal points with commas, and dropped or added minus signs. It can remove leading zeros from account numbers, product codes, and postal codes.
• Store identifiers and values requiring exact formatting as text. Dates may also be reinterpreted by spreadsheet software according to regional settings.
• Use CSV UTF-8 for accented or non-Latin characters, and verify the delimiter, line-ending convention, and quote handling expected by the receiving system.
• Open the saved CSV in a plain-text editor or re-import it into the destination application to confirm that columns, line breaks, character encoding, and quoted fields remain correct.