Recognition Presets

Recognition presets

Aspose.BarCode for Java provides several recognition presets through QualitySettings.
These presets bundle multiple internal options (deconvolution, noise filters, search depth, etc.) to give you a convenient speed vs robustness trade-off:

  • QualitySettings.getHighPerformance() – favor speed on clean, high-quality images.
  • QualitySettings.getHighQuality() – enable additional recovery methods for degraded barcodes.
  • QualitySettings.getMaxQuality() – use the most exhaustive, slowest analysis for the toughest cases.

This article demonstrates how to apply these presets in practice based on the example class:

com.aspose.barcode.guide.recognition.performance.RecognitionPresets

You can find the full source code on GitHub:

RecognitionPresets.java

All snippets below assume that helper methods such as ExampleAssist.checkOrCreateImage(...) and ExampleAssist.assertRecognized(...) are available in your project.

1. Test data: clean vs degraded Code 128

The example class uses two types of barcode images:

  • Clean Code 128 – simple, high-contrast symbol generated at a comfortable size.
  • Blurred + noisy Code 128 – the same payload rendered larger and then degraded with blur and salt-and-pepper noise.

The clean image is generated as follows:

private void generateCode128Clean(String fullPath) throws IOException {
    BarcodeGenerator barcodeGenerator = new BarcodeGenerator(EncodeTypes.CODE_128, "SPEED-QUALITY-DEMO");
    barcodeGenerator.getParameters().getBarcode().getXDimension().setPixels(2);
    barcodeGenerator.getParameters().getBarcode().getBarHeight().setPixels(60);
    barcodeGenerator.save(fullPath, BarCodeImageFormat.PNG);
}

The degraded variant uses a custom blur + noise pipeline (simplified):

private void generateCode128BlurredNoisy(String fullPath) throws IOException {
    // 1) Render a larger, high-quality Code 128
    String hi = fullPath + ".hi.png";
    BarcodeGenerator barcodeGenerator = new BarcodeGenerator(EncodeTypes.CODE_128, "ROBUSTNESS-CHECK-123");
    barcodeGenerator.getParameters().getBarcode().getXDimension().setPixels(6);   // 3 * scale
    barcodeGenerator.getParameters().getBarcode().getBarHeight().setPixels(140);  // 70 * scale
    // ... configure generous quiet zones ...
    barcodeGenerator.save(hi, BarCodeImageFormat.PNG);

    // 2) Apply a light 3×3 box blur and salt & pepper noise
    BufferedImage src = ImageIO.read(Paths.get(hi).toFile());
    BufferedImage blurred = boxBlur(src);
    addSaltPepperNoise(blurred, 0.004);

    // 3) Downscale back to target size (bilinear interpolation)
    //    and save the final degraded image
    // ...
}

You do not need to replicate this degradation logic in your application; it is only used to produce deterministic test fixtures for comparing presets.

2. HighPerformance on clean barcodes

Use HighPerformance when:

  • you expect clean, high-quality barcodes (for example, generated labels),
  • you want to prioritize throughput and low CPU usage,
  • occasional difficult edge cases can be handled separately with more robust settings.

Example: fast recognition of a clean Code 128 symbol.

String fileName = "code128_clean.png";

ExampleAssist.checkOrCreateImage(IMAGES_FOLDER, fileName, this::generateCode128Clean);

QualitySettings qualitySettings = QualitySettings.getHighPerformance();
BarCodeReader barCodeReader = new BarCodeReader(
        Paths.get(IMAGES_FOLDER, fileName).toString(),
        DecodeType.CODE_128
);
barCodeReader.setQualitySettings(qualitySettings);

ExampleAssist.assertRecognized(barCodeReader, fileName, 1, DecodeType.CODE_128);

Key points:

  • Start from QualitySettings.getHighPerformance().
  • Pass this instance into barCodeReader.setQualitySettings(...) before calling readBarCodes() or helper methods.
  • On clean input, this preset should decode quickly while using fewer resources.

3. MaxQuality on clean images (thorough analysis)

MaxQuality enables the most exhaustive set of recognition procedures.
Even though it is not required for clean images, running it on the same test input helps illustrate the trade-off: higher robustness potential vs longer processing time.

Example: using MaxQuality on a similar clean Code 128 image.

String fileName = "code128_max_quality.png";

ExampleAssist.checkOrCreateImage(IMAGES_FOLDER, fileName, this::generateCode128Clean);

QualitySettings qualitySettings = QualitySettings.getMaxQuality();
BarCodeReader barCodeReader = new BarCodeReader(
        Paths.get(IMAGES_FOLDER, fileName).toString(),
        DecodeType.CODE_128
);
barCodeReader.setQualitySettings(qualitySettings);

ExampleAssist.assertRecognized(barCodeReader, fileName, 1, DecodeType.CODE_128);

Recommendations:

  • Reserve MaxQuality for challenging inputs (poor lighting, severe distortions, difficult backgrounds).
  • For normal production flow, prefer NormalQuality or HighQuality and use MaxQuality in a fallback path or for re-processing.

4. HighQuality on blurred and noisy barcodes

HighQuality presets are targeted at degraded images: slight blur, noise, or poor contrast.
In the example, a Code 128 image is blurred, sprinkled with salt-and-pepper noise, and then downscaled. The HighQuality preset is used to recover it.

String fileName = "code128_blur_noise.png";

ExampleAssist.checkOrCreateImage(IMAGES_FOLDER, fileName, this::generateCode128BlurredNoisy);

QualitySettings qualitySettings = QualitySettings.getHighQuality();
BarCodeReader barCodeReader = new BarCodeReader(
        Paths.get(IMAGES_FOLDER, fileName).toString(),
        DecodeType.CODE_128
);
barCodeReader.setQualitySettings(qualitySettings);

ExampleAssist.assertRecognized(barCodeReader, fileName, 1, DecodeType.CODE_128);

Why HighQuality helps on damaged inputs:

  • it may enable more aggressive deconvolution and noise suppression,
  • it increases the search space for potential barcode candidates,
  • it can apply additional robustness filters that would be too expensive for realtime-only scenarios.

Use this preset when you cannot fully control input quality (for example, mobile cameras or scanned documents).

5. Measuring speed vs quality trade-off

The example class also contains a method that compares elapsed time for HighPerformance vs MaxQuality on the same image.
This test does not enforce any strict timing; it only prints measured values so you can estimate the relative cost.

String fileName = "compare_timing_high_quality.png";

ExampleAssist.checkOrCreateImage(IMAGES_FOLDER, fileName, this::generateCode128Clean);

// HighPerformance timing
BarCodeReader fastReader = new BarCodeReader(path(fileName), DecodeType.CODE_128);
fastReader.setQualitySettings(QualitySettings.getHighPerformance());
long t1 = System.nanoTime();
ExampleAssist.assertRecognized(fastReader, "HighPerformance timing", 1, DecodeType.CODE_128);
long fastMs = (System.nanoTime() - t1) / 1_000_000L;

// MaxQuality timing
BarCodeReader thoroughReader = new BarCodeReader(path(fileName), DecodeType.CODE_128);
thoroughReader.setQualitySettings(QualitySettings.getMaxQuality());
long t2 = System.nanoTime();
ExampleAssist.assertRecognized(thoroughReader, "MaxQuality timing", 1, DecodeType.CODE_128);
long thoroughMs = (System.nanoTime() - t2) / 1_000_000L;

System.out.println(
        "[Speed vs Quality] HighPerformance: " + fastMs + " ms; MaxQuality: " + thoroughMs + " ms"
);

How to use such measurements in your project:

  • Run similar tests on your own sample set (typical resolutions, camera devices, printers).
  • Log both recognition time and success rate (how many barcodes fail to decode).
  • Choose a preset (or combination of presets) that satisfies your latency budget while keeping failure rate acceptable.

6. Practical recommendations

When configuring recognition in a real application, a common strategy is:

  1. Start fast
    Use HighPerformance or NormalQuality for the first pass on incoming frames or images.

  2. Fallback on failure
    If no barcodes are recognized (or confidence is too low), re-run recognition on the same image with a more robust preset such as HighQuality or MaxQuality.

  3. Tune per scenario

    • For scanned documents or images captured in controlled conditions, NormalQuality or HighPerformance is often enough.
    • For mobile camera input (low light, perspective distortions, motion blur), consider HighQuality or a dedicated slow path for re-processing.

  4. Measure and iterate
    Use tests similar to compareTimingHighPerformanceVsMaxQuality() to collect timing statistics and make data-driven decisions about presets.

All examples in this article are based on:

RecognitionPresets.java

Use them as a template for integrating recognition presets into your own Java applications.

Reading Low-Resolution Barcode