OCR chính xác: Cải thiện hình ảnh chất lượng thấp trong .NET

dotnet new console -n OCRImprovementExample -f net8.0
cd OCRImprovementExample
dotnet add package Aspose.OCR

using System;
using System.Drawing;
using System.IO;

// Step 1: Load the low-quality image
string imagePath = "low_quality_image.png";
if (!File.Exists(imagePath))
    throw new FileNotFoundException(imagePath);

Bitmap image = (Bitmap)Image.FromFile(imagePath);

// Optional: quick sanity check
Console.WriteLine($"Loaded {imagePath} ({image.Width}x{image.Height}px)");

using Aspose.Ocr.ImageProcessing;

// Median filter: try odd sizes 3, 5 (larger = stronger, but may blur small text)
var filterOptions = new FilterOptions
{
    MedianFilter = true,
    MedianFilterSize = 3
};

var contrastOptions = new ContrastOptions
{
    // Positive values increase contrast/brightness; negatives decrease
    ContrastAdjustment = 20,   // try 10..30
    BrightnessAdjustment = 10  // try -10..+15 based on exposure
};

var processingOptions = new ImageProcessingOptions();
processingOptions.Filters.Add(filterOptions);
processingOptions.Contrast = contrastOptions;

// (Optional) more options can be added here if your build exposes them
// e.g., processingOptions.Sharpen = new SharpenOptions { Strength = 1 };

using (var ocrEngine = new Aspose.Ocr.Api.OcrEngine())
{
    Bitmap preprocessed = ocrEngine.PreprocessImage(image, processingOptions);
    // Keep this for OCR below
    image.Dispose();
    image = preprocessed;
}

// 1.5 Optional: upscale 1.5x to help recognition of tiny text
Bitmap Upscale(Bitmap src, double scale)
{
    int w = (int)(src.Width * scale);
    int h = (int)(src.Height * scale);
    var dest = new Bitmap(w, h);
    using (var g = Graphics.FromImage(dest))
    {
        g.InterpolationMode = System.Drawing.Drawing2D.InterpolationMode.HighQualityBicubic;
        g.DrawImage(src, 0, 0, w, h);
    }
    return dest;
}

// Example usage
// image = Upscale(image, 1.5);

// Crop a rectangle (x,y,width,height)
Rectangle roi = new Rectangle(0, 0, image.Width, Math.Min(400, image.Height)); // top band
Bitmap cropped = image.Clone(roi, image.PixelFormat);
image.Dispose();
image = cropped;

// Simple grayscale + global threshold (0..255); try 170..200
Bitmap ToBinary(Bitmap src, byte threshold = 185)
{
    var bw = new Bitmap(src.Width, src.Height);
    for (int y = 0; y < src.Height; y++)
    for (int x = 0; x < src.Width; x++)
    {
        var c = src.GetPixel(x, y);
        byte gray = (byte)(0.299 * c.R + 0.587 * c.G + 0.114 * c.B);
        byte v = gray >= threshold ? (byte)255 : (byte)0;
        bw.SetPixel(x, y, Color.FromArgb(v, v, v));
    }
    return bw;
}

// Example usage
// var bin = ToBinary(image, 190);
// image.Dispose();
// image = bin;

// Only if your build exposes these settings:
using Aspose.Ocr;

var settingsAvailable = false; // flip true if your API supports it
// Example (may vary by package version):
// ocrEngine.Settings.Language = RecognitionLanguages.English;
// ocrEngine.Settings.PageSegmentationMode = PageSegmentationMode.Auto;

using System;
using System.Drawing;
using Aspose.Ocr;
using Aspose.Ocr.ImageProcessing;

namespace OCRImprovementExample
{
    class Program
    {
        static void Main(string[] args)
        {
            string imagePath = "low_quality_image.png";
            Bitmap image = (Bitmap)Image.FromFile(imagePath);

            // Preprocess (median + contrast/brightness)
            var filterOptions = new FilterOptions { MedianFilter = true, MedianFilterSize = 3 };
            var contrastOptions = new ContrastOptions { ContrastAdjustment = 20, BrightnessAdjustment = 10 };
            var processingOptions = new ImageProcessingOptions();
            processingOptions.Filters.Add(filterOptions);
            processingOptions.Contrast = contrastOptions;

            using (Aspose.Ocr.Api.OcrEngine ocrEngine = new Aspose.Ocr.Api.OcrEngine())
            {
                // Preprocess
                Bitmap preprocessedImage = ocrEngine.PreprocessImage(image, processingOptions);

                // OCR
                string recognizedText = ocrEngine.RecognizeImage(preprocessedImage);

                Console.WriteLine("Recognized Text:");
                Console.WriteLine(recognizedText);
            }
        }
    }
}

File.WriteAllText("recognized.txt", recognizedText);