thesis-src/Examples/Common/ScreenshotScanner.cs

using ImageMagick;
using Lookup.Memory;
using Ocr.Tesseract;
using Ocr.Tesseract.Configuration;
using Ocr.Tesseract.Models;
using Ocr.Tesseract.Screenshots;
using Ocr.Tesseract.Screenshots.Configuration;
using Ocr.Tesseract.Screenshots.Threshold;
using Process.Abstract.Configuration;
using Process.Interface;
using System.Text.RegularExpressions;

namespace Common
{
  /// <summary>
  /// Scanner class, scanning <see cref="MagickImage"/>s for <see cref="Word"/>s
  /// via optical character recognition. Optimized for digital Screenshots.
  /// </summary>
  public class ScreenshotScanner
  {
    private readonly IProcessor<MagickImage, ScanResult> _processor;

    /// <summary>
    /// <see cref="Regex"/> expression for extracting whole words from scan results
    /// </summary>
    private static readonly Regex wordRegex = new(
      @"[\w'\-]{2,}",
      RegexOptions.Compiled | RegexOptions.Multiline | RegexOptions.IgnoreCase
    );

    /// <summary>
    /// Data storage
    /// </summary>
    public Lookup.Interface.ILookup<Word, MagickImage> Lookup { get; } =
      new MemoryLookup<Word, MagickImage>();

    /// <summary>
    /// Configuration of the <see cref="ImageProcessor"/>
    /// </summary>
    public ScreenshotProcessorConfiguration ImageProcessorConfiguration { get; }

    public ITesseractConfiguration TesseractConfiguration { get; }

    /// <summary>
    /// Constructor
    /// </summary>
    public ScreenshotScanner(
      ScreenshotProcessorConfiguration imageProcessorConfig,
      ITesseractConfiguration tesseractConfig
    )
    {
      ImageProcessorConfiguration = imageProcessorConfig;
      TesseractConfiguration = tesseractConfig;

      _processor = MakeProcessor();
    }

    /// <summary>
    /// Process the provided <paramref name="images"/> and add the results to
    /// the <see cref="Lookup"/>
    /// </summary>
    /// <param name="images">The <see cref="MagickImage"/>s to process</param>
    public void Process(IEnumerable<MagickImage> images)
    {
      foreach (var kv in _processor.Process(images))
      {
        Lookup.Add(kv.Word, kv.Image);
      }
    }

    private IProcessor<MagickImage, ScanResult> MakeProcessor()
    {
      var threshold =
        new ThresholdAdaptiveProcessor(ImageProcessorConfiguration.ThresholdWidth,
          ImageProcessorConfiguration.ThresholdHeight);
      // var threshold = new AutoThresholdProcessor(AutoThresholdMethod.Kapur);
      // var threshold = new AutoThresholdProcessor(AutoThresholdMethod.OTSU);
      // var threshold = new AutoThresholdProcessor(AutoThresholdMethod.Triangle);
      // var threshold = new ThresholdProcessor(60);

      var preprocessing = new ProcessorChainConfiguration<MagickImage, MagickImage>()
        .Use(new CloneImageProcessor())
        .Use(new ResizeProcessor(FilterType.Lanczos2Sharp, PixelInterpolateMethod.Mesh))
        .Use(new NormalizeProcessor())
        .Use(threshold)
        .Use(new AddBorderProcessor(10))
        .Use(new BinarizeProcessor())
        .Complete(new NegateCloneProcessor());

      var postprocessing = new ProcessorChainConfiguration<ScanResult, ScanResult>()
        .Use(new ConfidenceFilter(50))
        .Use(new ToLowerProcessor())
        .Use(new DuplicateFilter())
        .Complete(new RegexFilter(wordRegex));

      var scan = new TesseractProcessor(TesseractConfiguration);

      return new ProcessorChainConfiguration<MagickImage, ScanResult>()
        .Use(preprocessing)
        .Use(new ProcessingEvent<MagickImage>(OnProcessing))
        .Use(scan)
        .Use(new ProcessingEvent<ScanResult>(OnProcessed))
        .Complete(postprocessing);
    }

    protected virtual void OnProcessing(IProcessor sender, ICollection<MagickImage> inputs)
    {
    }

    protected virtual void OnProcessed(IProcessor sender, ICollection<ScanResult> inputs)
    {
    }
  }
}