EasyOCR Documentation

EasyOCR is a Java-based OCR (Optical Character Recognition) engine built upon Tesseract. It is designed primarily for developers and provides localized SDK integration for C/S (Client/Server), B/S (Browser/Server), and Android mobile projects.

Key features include:

• Image Processing: Integrated image cleanup capabilities.
• Specialized Recognition: Support for CAPTCHA recognition and bill/note processing.
• Extensibility: Supports plugin programming and ETD (Electronic Template Document) templates.
• GUI Tools: Provides EasyTemplateDesigner GUI for graphical ETD template design.

Note on Licensing: As of version 4.X, EasyOCR is no longer open source or free. Commercial users can access specific documentation for development, installation (Windows/Non-Windows), API usage, plugin development, and Java docs.

To use EasyOCR in your Java project, follow these three steps:

1. Install Engine: Set up the underlying OCR engine.
2. Add the jar package: Include the EasyOCR library in your project dependencies.
3. Call API: Use the EasyOCR class to perform recognition tasks.

EasyOCR e = new EasyOCR();
// Direct recognition of image content
System.out.println(e.recognize("images/demo_eurotext.png"));

To perform simple OCR on an image, instantiate the EasyOCR class and call the recognize method with the image path. This method handles the core OCR process in a single line of code.

EasyOCR e = new EasyOCR();
// Directly recognize image content
System.out.println(e.recognize("images/demo_eurotext.png")); 

Use the static method EasyOCR.findTextRegions to locate specific text areas within a BufferedImage. This method returns a list of TextRegion objects containing coordinates.

String filePath="./images/idcard.png";
BufferedImage image = ImageIO.read(new File(filePath));

// Find Text Regions with specific parameter bounds
List<TextRegion> regions = EasyOCR.findTextRegions(image, 10, 20, 70, 200);

// 'regions' contains TextRegion objects with x, y, x2, y2 coordinates

EasyOCR provides several key classes for different OCR tasks:

• EasyOCR: The primary class for text recognition. It handles the full lifecycle including automatic image cleanup and recognition integration. It supports template-based recognition and cleaning.
• ImageClean: Used for cleaning images, specifically optimized for CAPTCHAs and notes. It supports various predefined cleanup modes to handle deformation and rotation to improve recognition rates.
• Language: A list of supported languages. Supports simultaneous mixed-language recognition (e.g., English + Japanese).
• TextMode: An enumeration for selecting different text recognition modes.
• CleanType: An enumeration for image filtering types, including NONE, CAPTCHA (verification code cleanup), TEXT (text cleaning), and bill image cleanup algorithms.

You can detect specific text regions within a BufferedImage using the static method EasyOCR.findTextRegions. This returns a list of TextRegion objects containing coordinates.

String filePath="./images/idcard.png";
BufferedImage image = ImageIO.read(new File(filePath));

// Find Text Regions with custom bounding parameters
List<TextRegion> regions = EasyOCR.findTextRegions(image, 10, 20, 70, 200);

for (TextRegion r : regions) {
    // Process each region (e.g., draw rectangles around them)
    System.out.println("Region found at: " + r.x + ", " + r.y);
}

The ImageType (referred to as CleanType in some contexts) defines the specific cleaning and recognition algorithms used for different image types.

- CAPTCHA_NORMAL: Normal CAPTCHA images
- CAPTCHA_INTERFERENCE_LINE: CAPTCHA images with interference lines
- CAPTCHA_SPOT: Dot-based CAPTCHA images
- CAPTCHA_WHITE_CHAR: White text on solid background CAPTCHA
- CAPTCHA_HOLLOW_CHAR: Hollow character CAPTCHA
- CLEAR: General image clarification (no special interference)
- LINK_BOLD: Bold/connected fonts
- BILL_NORMAL: Normal bill/invoice type
- TEXT_BOLD_BLAK: Bold black text

The following classes form the core API of EasyOCR:

- EasyOCR: Core class for OCR engine calls. Supports recognize(), recognizeAutoCleanImage(), and recognizeByTemplate().
- ImageClean: Class for cleaning CAPTCHAs, bills, and regular images.
- Language: List of supported recognition languages (supports mixed language).
- TextMode: Enumeration of text recognition modes.
- CleanType: Enumeration for image filtering/cleaning types (NONE, CAPTCHA, TEXT, etc.).