在 .NET 中检测文本文档图像的方向

VintaSoft Imaging .NET SDK 和 VintaSoft Document Cleanup .NET Plug-in 提供用于处理和清理文档图像的命令。
GetTextOrientationCommand命令的目的是检测包含旋转90度、180度或270度拉丁文本的图像的方向。
GetTextOrientationCommand 不适用于处理：

• 不包含文本的图像
• 包含非拉丁文本的图像
• 图像，仅包含大写字母的拉丁文本
• 图像，其旋转角度不是 0、90、180 或 270 度（使用 DeskewCommand 命令在检测图像方向之前对其进行倾斜校正）。
• 图像，仅包含少量单词（这不足以判断图像方向）

VintaSoft Imaging .NET SDK 和 VintaSoft OCR .NET Plug-in 提供了 GetTesseractOcrTextOrientationCommand 命令，其目的是使用 Tesseract OCR 引擎检测文档图像的方向。 GetTesseractOcrTextOrientationCommand 命令可以检测任何包含文本的图像的方向（不像 GetTextOrientationCommand 命令那样有限制），但是GetTesseractOcrTextOrientationCommand 命令的运行速度比 GetTextOrientationCommand 命令慢 5 倍。


基于上述情况，需要使用 GetTextOrientationCommand 命令或 GetTesseractOcrTextOrientationCommand 命令取决于输入文档图像的类型。

也可以同时使用这两个命令，具体如下：

• 使用GetTextOrientationCommand命令检测图像方向（速度更快，并且对于大多数包含拉丁文本的文档，检测方向准确）
• 如果 GetTextOrientationCommand 命令无法检测文档图像的方向，则使用 GetTesseractOcrTextOrientationCommand 命令检测图像方向（速度较慢，但​​几乎可以正确检测所有文档的方向）。

结合使用这两个命令可以最大限度地提高文档图像方向检测的效率和质量。


以下是 C# 代码，演示了如何使用 GetTextOrientationCommand 命令检测文档图像的方向：

/// <summary>
/// Returns an orientation angle of document image using statistics for Latin symbols.
/// </summary>
/// <param name="filename">The path to a file with document image.</param>
public static void GetDocumentImageOrientationUsingLatinSymbolStat(string filename)
{
    // create an image collection
    using (Vintasoft.Imaging.ImageCollection images = new Vintasoft.Imaging.ImageCollection())
    {
        // add images from file to the image collection
        images.Add(filename);

        // create an instance of GetTextOrientationCommand class
        Vintasoft.Imaging.ImageProcessing.Info.GetTextOrientationCommand getTextOrientationCommand1 =
            new Vintasoft.Imaging.ImageProcessing.Info.GetTextOrientationCommand();

        // for each image in image collection
        for (int i = 0; i < images.Count; i++)
        {
            // get image
            Vintasoft.Imaging.VintasoftImage image = images[i];

            // determine orientation of document image using statistics for Latin symbols
            getTextOrientationCommand1.ExecuteInPlace(image);

            // write result to the console
            System.Console.WriteLine(string.Format("Filename: {0}, page: {1}, page orientation: {2}, confidence: {3}",
                System.IO.Path.GetFileName(filename),
                i,
                getTextOrientationCommand1.Orientation,
                getTextOrientationCommand1.Confidence));
        }

        // free images
        images.ClearAndDisposeItems();
    }
}

以下是 C# 代码，演示了如何使用 GetTesseractOcrTextOrientationCommand 命令：

/// <summary>
/// Returns an orientation angle of document image using Tesseract OCR.
/// </summary>
/// <param name="filename">The path to a file with document image.</param>
/// <param name="tesseractOcrDllDirectory">A path to a directory, where Tesseract5.Vintasoft.xXX.dll files are located.</param>
public static void GetDocumentImageOrientationUsingTesseractOCR(string filename, string tesseractOcrDllDirectory)
{
    // create an image collection
    using (Vintasoft.Imaging.ImageCollection images = new Vintasoft.Imaging.ImageCollection())
    {
        // add images from file to the image collection
        images.Add(filename);

        // create an instance of GetTesseractOcrTextOrientationCommand class
        using (Vintasoft.Imaging.ImageProcessing.Ocr.Tesseract.GetTesseractOcrTextOrientationCommand getTextOrientationCommand =
            new Vintasoft.Imaging.ImageProcessing.Ocr.Tesseract.GetTesseractOcrTextOrientationCommand())
        {
            // specify path to a directory, where Tesseract5.Vintasoft.xXX.dll files are located
            getTextOrientationCommand.TesseractOcrDllDirectory = tesseractOcrDllDirectory;

            // for each image in image collection
            for (int i = 0; i < images.Count; i++)
            {
                // get image
                Vintasoft.Imaging.VintasoftImage image = images[i];

                // determine orientation of document image using Tesseract OCR
                getTextOrientationCommand.ExecuteInPlace(image);

                // write result to the console
                System.Console.WriteLine(string.Format("Filename: {0}, page: {1}, page orientation: {2}",
                    System.IO.Path.GetFileName(filename),
                    i,
                    getTextOrientationCommand.Orientation));
            }
        }

        // free images
        images.ClearAndDisposeItems();
    }
}

以下是 C# 代码，这演示了如何使用 GetTextOrientationCommand 命令和 GetTesseractOcrTextOrientationCommand 命令检测文档图像的方向：

/// <summary>
/// Returns an orientation angle of document image using statistics for Latin symbols and using Tesseract OCR.
/// </summary>
/// <param name="filename">The path to a file with document image.</param>
/// <param name="tesseractOcrDllDirectory">A path to a directory, where Tesseract5.Vintasoft.xXX.dll files are located.</param>
public static void GetDocumentImageOrientationUsingLatinSymbolStatAndOcrTesseract(string filename, string tesseractOcrDllDirectory)
{
    // create an image collection
    using (Vintasoft.Imaging.ImageCollection images = new Vintasoft.Imaging.ImageCollection())
    {
        // add images from file to the image collection
        images.Add(filename);

        // create an instance of GetTextOrientationCommand class
        Vintasoft.Imaging.ImageProcessing.Info.GetTextOrientationCommand getTextOrientationCommand1 =
            new Vintasoft.Imaging.ImageProcessing.Info.GetTextOrientationCommand();

        // create an instance of GetTesseractOcrTextOrientationCommand class
        using (Vintasoft.Imaging.ImageProcessing.Ocr.Tesseract.GetTesseractOcrTextOrientationCommand getTextOrientationCommand2 =
            new Vintasoft.Imaging.ImageProcessing.Ocr.Tesseract.GetTesseractOcrTextOrientationCommand())
        {
            // specify path to a directory, where Tesseract5.Vintasoft.xXX.dll files are located
            getTextOrientationCommand2.TesseractOcrDllDirectory = tesseractOcrDllDirectory;

            // for each image in image collection
            for (int i = 0; i < images.Count; i++)
            {
                // get image
                Vintasoft.Imaging.VintasoftImage image = images[i];

                // determine orientation of document image using statistics for Latin symbols (works for Latin text only)
                getTextOrientationCommand1.ExecuteInPlace(image);
                // if orientation is detected and orientation result has high confidence
                if (getTextOrientationCommand1.Orientation != Vintasoft.Imaging.ImageProcessing.Info.ImageOrthogonalOrientation.Undefined &&
                    getTextOrientationCommand1.Confidence > 0.3)
                {
                    // write result to the console
                    System.Console.WriteLine(string.Format("Filename: {0}, page: {1}, page orientation: {2}, confidence: {3}",
                        System.IO.Path.GetFileName(filename),
                        i,
                        getTextOrientationCommand1.Orientation,
                        getTextOrientationCommand1.Confidence));
                }
                // if orientation is not detected or orientation result has low confidence
                else
                {
                    // determine orientation of document image using Tesseract OCR (works for any text)
                    getTextOrientationCommand2.ExecuteInPlace(image);

                    // write result to the console
                    System.Console.WriteLine(string.Format("Filename: {0}, page: {1}, page orientation: {2}, confidence: {3}",
                        System.IO.Path.GetFileName(filename),
                        i,
                        getTextOrientationCommand2.Orientation,
                        getTextOrientationCommand2.Confidence));
                }
            }
        }

        // free images
        images.ClearAndDisposeItems();
    }
}