Get text preceding image in PDF

问题:

What I'm trying to do is extract the image associated with some text in a PDF file. For instance, a PDF would have a photo of the front of a house. Just above the photo, there would be a caption which reads "Front View". I want the program to search the PDF for the text "Front View" and extract the photo that follows it.

I've looked iTextSharp, PDFsharp, and other utilities, but all of them treat the text in a PDF and the images separately. There doesn't seem to be any way to figure out that this line of text comes before that image.

We use iTextSharp for manipulating PDFs. I've written a method in C# that will extract an image given a page number, the number of the image on the page, and the image type. For instance, I can extract the 2nd jpeg on page 3. Here is the code for that. What I would like is to be able to search for a line of text in the file and then extract the image that follows that line of text.

public class ImageExtractor : IRenderListener
{
    int _currentPage = 1;
    int _imageCount = 0;
    int _index = 0;
    int _count = 0;
    readonly string _outputFilePrefix;
    readonly string _outputFolder;
    readonly bool _overwriteExistingFiles;
    string[] _fileTypes;

    public ImageExtractor(string outputFilePrefix, string outputFolder, bool overwriteExistingFiles, string[] fileTypes, int index)
    {
        _outputFilePrefix = outputFilePrefix;
        _outputFolder = outputFolder;
        _overwriteExistingFiles = overwriteExistingFiles;
        _fileTypes = fileTypes;
        _index = index;
    }

    public static int ExtractImageByIndex(string pdfPath, string outputFilePrefix, string outputFolder, bool overwriteExistingFiles, int pageNumber, int index, string[] fileTypes = null)
    {
        // Handle setting of any default values
        outputFilePrefix = outputFilePrefix ?? System.IO.Path.GetFileNameWithoutExtension(pdfPath);
        outputFolder = String.IsNullOrEmpty(outputFolder) ? System.IO.Path.GetDirectoryName(pdfPath) : outputFolder;

        var instance = new ImageExtractor(outputFilePrefix, outputFolder, overwriteExistingFiles, fileTypes, index);
        instance._currentPage = pageNumber;

        using (var pdfReader = new PdfReader(pdfPath))
        {
            if (pdfReader.NumberOfPages == 0)
                return 0;

            if (pdfReader.IsEncrypted())
                throw new ApplicationException(pdfPath + " is encrypted.");

            var pdfParser = new PdfReaderContentParser(pdfReader);

            pdfParser.ProcessContent(instance._currentPage, instance);
        }

        return instance._imageCount;
    }

    public void BeginTextBlock() { }
    public void EndTextBlock() { }
    public void RenderText(TextRenderInfo renderInfo) { }

    public void RenderImage(ImageRenderInfo renderInfo)
    {
        // If _index is greater than 0, we're looking for a specific image. If _count is
        // equal to _index, we've already found it, so don't go any farther.
        if (_index > 0 && _count == _index)
            return;

        var imageObject = renderInfo.GetImage();

        var imageFileName = "";

        if (_fileTypes != null)
        {
            var type = imageObject.GetFileType().ToLower();
            var flag = false;
            foreach (var t in _fileTypes)
            {
                if (t.ToLower() == type)
                {
                    flag = true;
                    break;
                }
            }
            if (flag)
                imageFileName = String.Format("{0}_{1}_{2}.{3}", _outputFilePrefix, _currentPage, _imageCount, imageObject.GetFileType());
        }
        else
        {
            imageFileName = String.Format("{0}_{1}_{2}.{3}", _outputFilePrefix, _currentPage, _imageCount, imageObject.GetFileType());
        }

        if (!string.IsNullOrEmpty(imageFileName))
        {
            // If _index is 0, multiple images may be extracted. If _index is greater than 0,
            // RenderImage will increment count every time it finds an image that matches the
            // file type and will only extract the image if count equals index.
            if (_index > 0)
            {
                _count++;
                if (_count != _index)
                    return;
            }

            var imagePath = System.IO.Path.Combine(_outputFolder, imageFileName);

            if (_overwriteExistingFiles || !File.Exists(imagePath))
            {
                var imageRawBytes = imageObject.GetImageAsBytes();

                File.WriteAllBytes(imagePath, imageRawBytes);

            }

            // Subtle: Always increment even if file is not written. This ensures consistency should only some
            //   of a PDF file's images actually exist.
            _imageCount++;
        }
    }
}

回答1:

As already mentioned in a comment, this is very similar to the topic of the question Extraction of images present inside a paragraph with the main difference that in the context of that question iText for Java was used instead of iTextSharp for .Net.

A port of the Java SimpleMixedExtractionStrategy from that question might look like this:

public class SimpleMixedExtractionStrategy : LocationTextExtractionStrategy
{
    FieldInfo field = typeof(LocationTextExtractionStrategy).GetField("locationalResult", BindingFlags.Instance | BindingFlags.NonPublic);
    LineSegment UNIT_LINE = new LineSegment(new Vector(0, 0, 1), new Vector(1, 0, 1));
    String outputPath;
    String name;
    int counter = 0;

    public SimpleMixedExtractionStrategy(String outputPath, String name)
    {
        this.outputPath = outputPath;
        this.name = name;
    }

    public override void RenderImage(ImageRenderInfo renderInfo)
    {
        PdfImageObject image = renderInfo.GetImage();
        if (image == null) return;
        int number = counter++;
        String filename = name + "-" + number + "." + image.GetFileType();
        File.WriteAllBytes(outputPath + filename, image.GetImageAsBytes());

        LineSegment segment = UNIT_LINE.TransformBy(renderInfo.GetImageCTM());
        TextChunk location = new TextChunk("[" + filename + "]", segment.GetStartPoint(), segment.GetEndPoint(), 0f);

        List<TextChunk> locationalResult = (List<TextChunk>)field.GetValue(this);
        locationalResult.Add(location);
    }
}

Just like in the Java implementation, it is necessary to use reflection to access the private List<TextChunk> locationalResult in LocationTextExtractionStrategy. If the use of reflection is not allowed in your project, you can copy the whole source of LocationTextExtractionStrategy to an own class and apply the changes to the copy.

You might use it like this:

String sourceFile = @"SOURCE.pdf";
String imagePath = @"extract\";
String imageBaseName = "SOURCE-";
Directory.CreateDirectory(imagePath);

using (PdfReader pdfReader = new PdfReader(sourceFile))
{
    PdfReaderContentParser parser = new PdfReaderContentParser(pdfReader);
    for (var i = 1; i <= pdfReader.NumberOfPages; i++)
    {
        SimpleMixedExtractionStrategy listener = new SimpleMixedExtractionStrategy(imagePath, imageBaseName + i);
        parser.ProcessContent(i, listener);
        String text = listener.GetResultantText();
        Console.Write("Text of page {0}:\n---\n{1}\n\n", i, text);
    }
}

For the example file from the referred-to question

the output is:

Text of page 1:
---
Getting Started with Vaadin
• A version of Book of Vaadin that you can browse in the Eclipse Help system.
You can install the plugin as follows:
1. Start Eclipse.
2. Select Help   Software Updates....
3. Select the Available Software tab.
4. Add the Vaadin plugin update site by clicking Add Site....
[book-of-vaadin-page14-1-0.png]
Enter the URL of the Vaadin Update Site: http://vaadin.com/eclipse and click OK. The
Vaadin site should now appear in the Software Updates window.
5. Select all the Vaadin plugins in the tree.
[book-of-vaadin-page14-1-1.png]
Finally, click Install.
Detailed and up-to-date installation instructions for the Eclipse plugin can be found at http://vaad-
in.com/eclipse.
Updating the Vaadin Plugin
If you have automatic updates enabled in Eclipse (see Window   Preferences   Install/Update
  Automatic Updates), the Vaadin plugin will be updated automatically along with other plugins.
Otherwise, you can update the Vaadin plugin (there are actually multiple plugins) manually as
follows:
1. Select Help   Software Updates..., the Software Updates and Add-ons window will
open.
2. Select the Installed Software tab.
14 Vaadin Plugin for Eclipse

Thus, for your task

What I would like is to be able to search for a line of text in the file and then extract the image that follows that line of text.

simply search for that line of text in the output string above and look for the next line containing an image file name in square brackets.

(If your PDF also uses square brackets, you can envelop the file name in other delimiters in the SimpleMixedExtractionStrategy, e.g. some characters from a Unicode private use area.)

回答2:

Here's the solution I found. The original code contains a lot of stuff that's not directly related to the question, so I simplified it for the post.

public class ImageExtractor : IRenderListener
{
    private string caption;
    private bool _captionFound;
    private string _outputFolder;

    ....
    ....

    public void BeginTextBlock() { }

    public void EndTextBlock() { }

    public void RenderText(TextRenderInfo renderInfo) {
        // If this line of text contains the caption, set _captionFound to true
        if (renderInfo.GetText().Contains(_caption))
            _captionFound = true;
    }

    public void RenderImage(ImageRenderInfo renderInfo)
    {
        // Skip the image if _captionFound is false
        if (!_captionFound)
            return;

        // _captionFound is true, so extract the image

        // Code to extract image

        // Set _captionFound back to false, so that only the first image found is
        // extracted.
        _captionFound = false;

    }

    public static int ExtractImageByCaption(string caption, string pdfPath, string outputFolder, string outputFolder, bool overwriteExistingFiles, string[] fileTypes = null)
    {
        var instance = new ImageExtractor(outputFilePrefix, outputFolder, overwriteExistingFiles, fileTypes, 0);

        instance._caption = caption;
        instance._outputFolder = outputFolder;

        using (var pdfReader = new PdfReader(pdfPath))
        {
            if (pdfReader.IsEncrypted())
                throw new ApplicationException(pdfPath + " is encrypted.");

            var pdfParser = new PdfReaderContentParser(pdfReader);

            while (instance._currentPage <= pdfReader.NumberOfPages)
            {
                pdfParser.ProcessContent(instance._currentPage, instance);

                instance._currentPage++;
            }
        }
    }
}