For local OCR in Java, use Tess4J, a Java wrapper around the Tesseract OCR engine. Add the library and its runtime dependencies, provide Tesseract language data, set the path to that data, and call doOCR(...) on an image. For scanned PDFs, render each page as an image first; if a PDF already contains selectable text, extract that text directly instead of running OCR.
Read text from an image with Tess4J
Tess4J connects Java applications to Tesseract through JNA. Its documented image formats include TIFF, JPEG, GIF, PNG, and BMP, as well as multi-page TIFF workflows. The following example shows the basic API shape; it is illustrative, not a tested, version-pinned build.
import java.io.File;
import net.sourceforge.tess4j.ITesseract;
import net.sourceforge.tess4j.Tesseract;
import net.sourceforge.tess4j.TesseractException;
public class ImageTextReader {
public static String read(File image, String tessdataPath) throws TesseractException {
ITesseract ocr = new Tesseract();
ocr.setDatapath(tessdataPath);
return ocr.doOCR(image);
}
}Before calling read, add Tess4J and its runtime dependencies to the project, install the language files required for the text being recognized, and pass the directory containing those files as tessdataPath. The method returns recognized text as a string. Handle TesseractException in the calling application, and validate the result where incorrect text could affect decisions or downstream processing.
Prepare images for better OCR input
Tess4J’s guidance recommends at least 200 DPI and commonly around 300 DPI for OCR input. Monochrome or grayscale images are suitable starting points; use uncompressed TIFF or PNG when practical. PNG is lossless and often smaller, while TIFF can be useful for multi-image documents. Higher resolution by itself does not guarantee better recognition: font, skew, contrast, compression, language data, layout, and handwriting also affect results.
Choose the right workflow for a PDF
| PDF type | Recommended path | Why |
|---|---|---|
| Selectable-text PDF | Extract text with Apache PDFBox | The document already contains text, so OCR is usually unnecessary. |
| Scanned or image-only PDF | Use PDFBox to render or extract page images, then OCR each page with Tess4J/Tesseract | The pages are images rather than embedded text. |
PDFBox documents text extraction and image extraction operations, and Tess4J identifies PDFBox as its PDF-support path. Treat a multi-page scanned document as a page-by-page pipeline: render a page, send its image to OCR, then collect and validate the returned text.
Plan for deployment requirements
- Language data: Package or install the appropriate
tessdatafiles and configure the path your application uses. - Native runtime: Tess4J uses JNA to access Tesseract, so account for JNA, native libraries, image I/O support, and platform-specific binaries in the deployment environment.
- Versions: Pin Tess4J, Tesseract, PDFBox, and related native components to versions compatible with the project and target platforms. Compatibility and a tested dependency set are not established here.
- Licensing: Tesseract and Tess4J are released under Apache License 2.0. Review the relevant dependencies and distribution requirements for your application.
Know when this approach fits
Tess4J/Tesseract is a local, open-source OCR option when you want recognition within a Java application and can manage language files and native runtime components. Recognition quality is input-dependent, and no fixed accuracy percentage should be assumed. Cloud OCR is a different deployment and cost model; no provider comparison is established here.
Quick Recap
Best Value
Rank #4
Rank #2
References
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
