Converting PDF to multipage tiff (Group 4)

Question

I'm trying to convert PDFs as represented by the org.apache.pdfbox.pdmodel.PDDocument class and the icafe library (https://github.com/dragon66/icafe/) to a multipage tiff with group 4 compression and 300 dpi. The sample code works for me for 288 dpi but strangely NOT for 300 dpi, the exported tiff remains just white. Has anybody an idea what the issue is here?

The sample pdf which I use in the example is located here: http://www.bergophil.ch/a.pdf

import java.awt.image.BufferedImage;
import java.io.FileOutputStream;
import java.io.IOException;

import org.apache.pdfbox.pdmodel.PDDocument;
import org.apache.pdfbox.pdmodel.PDPage;

import cafe.image.ImageColorType;
import cafe.image.ImageParam;
import cafe.image.options.TIFFOptions;
import cafe.image.tiff.TIFFTweaker;
import cafe.image.tiff.TiffFieldEnum.Compression;
import cafe.io.FileCacheRandomAccessOutputStream;
import cafe.io.RandomAccessOutputStream;

public class Pdf2TiffConverter {
    public static void main(String[] args) {
        String pdf = "a.pdf";
        PDDocument pddoc = null;
        try {
            pddoc = PDDocument.load(pdf);
        } catch (IOException e) {
        }

        try {
            savePdfAsTiff(pddoc);
        } catch (IOException e) {
        }
    }

    private static void savePdfAsTiff(PDDocument pdf) throws IOException {
        BufferedImage[] images = new BufferedImage[pdf.getNumberOfPages()];
        for (int i = 0; i < images.length; i++) {
            PDPage page = (PDPage) pdf.getDocumentCatalog().getAllPages()
                    .get(i);
            BufferedImage image;
            try {
//              image = page.convertToImage(BufferedImage.TYPE_INT_RGB, 288); //works
                image = page.convertToImage(BufferedImage.TYPE_INT_RGB, 300); // does not work
                images[i] = image;
            } catch (IOException e) {
                e.printStackTrace();
            }
        }

        FileOutputStream fos = new FileOutputStream("a.tiff");
        RandomAccessOutputStream rout = new FileCacheRandomAccessOutputStream(
                fos);
        ImageParam.ImageParamBuilder builder = ImageParam.getBuilder();
        ImageParam[] param = new ImageParam[1];
        TIFFOptions tiffOptions = new TIFFOptions();
        tiffOptions.setTiffCompression(Compression.CCITTFAX4);
        builder.imageOptions(tiffOptions);
        builder.colorType(ImageColorType.BILEVEL);
        param[0] = builder.build();
        TIFFTweaker.writeMultipageTIFF(rout, param, images);
        rout.close();
        fos.close();
    }
}

Or is there another library to write multi-page TIFFs?

EDIT:

Thanks to dragon66 the bug in icafe is now fixed. In the meantime I experimented with other libraries and also with invoking ghostscript. As I think ghostscript is very reliable as id is a widely used tool, on the other hand I have to rely that the user of my code has an ghostscript-installation, something like this:

   /**
 * Converts a given pdf as specified by its path to an tiff using group 4 compression
 *
 * @param pdfFilePath The absolute path of the pdf
 * @param tiffFilePath The absolute path of the tiff to be created
 * @param dpi The resolution of the tiff
 * @throws MyException If the conversion fails
 */
private static void convertPdfToTiffGhostscript(String pdfFilePath, String tiffFilePath, int dpi) throws MyException {
    // location of gswin64c.exe
    String ghostscriptLoc = context.getGhostscriptLoc();

    // enclose src and dest. with quotes to avoid problems if the paths contain whitespaces
    pdfFilePath = "\"" + pdfFilePath + "\"";
    tiffFilePath = "\"" + tiffFilePath + "\"";

    logger.debug("invoking ghostscript to convert {} to {}", pdfFilePath, tiffFilePath);
    String cmd = ghostscriptLoc + " -dQUIET -dBATCH -o " + tiffFilePath + " -r" + dpi + " -sDEVICE=tiffg4 " + pdfFilePath;
    logger.debug("The following command will be invoked: {}", cmd);

    int exitVal = 0;
    try {
        exitVal = Runtime.getRuntime().exec(cmd).waitFor();
    } catch (Exception e) {
        logger.error("error while converting to tiff using ghostscript", e);
        throw new MyException(ErrorMessages.GHOSTSTSCRIPT_ERROR, e);
    }
    if (exitVal != 0) {
        logger.error("error while converting to tiff using ghostscript, exitval is {}", exitVal);
        throw new MyException(ErrorMessages.GHOSTSTSCRIPT_ERROR);
    }
}

I found that the produced tif from ghostscript strongly differs in quality from the tiff produced by icafe (the group 4 tiff from ghostscript looks greyscale-like)

Solution 1

It's been a while since the question was asked and I finally find time and a wonderful ordered dither matrix which allows me to give some details on how "icafe" can be used to get similar or better results than calling external ghostscript executable. Some new features were added to "icafe" recently such as better quantization and ordered dither algorithms which is used in the following example code.

Here the sample pdf I am going to use is princeCatalogue. Most of the following code is from the OP with some changes due to package name change and more ImageParam control settings.

import java.awt.image.BufferedImage;
import java.io.FileOutputStream;
import java.io.IOException;

import org.apache.pdfbox.pdmodel.PDDocument;
import org.apache.pdfbox.pdmodel.PDPage;

import com.icafe4j.image.ImageColorType;
import com.icafe4j.image.ImageParam;
import com.icafe4j.image.options.TIFFOptions;
import com.icafe4j.image.quant.DitherMethod;
import com.icafe4j.image.quant.DitherMatrix;
import com.icafe4j.image.tiff.TIFFTweaker;
import com.icafe4j.image.tiff.TiffFieldEnum.Compression;
import com.icafe4j.io.FileCacheRandomAccessOutputStream;
import com.icafe4j.io.RandomAccessOutputStream;

public class Pdf2TiffConverter {
    public static void main(String[] args) {
        String pdf = "princecatalogue.pdf";
        PDDocument pddoc = null;
        try {
            pddoc = PDDocument.load(pdf);
        } catch (IOException e) {
        }

        try {
            savePdfAsTiff(pddoc);
        } catch (IOException e) {
        }
    }

    private static void savePdfAsTiff(PDDocument pdf) throws IOException {
        BufferedImage[] images = new BufferedImage[pdf.getNumberOfPages()];
        for (int i = 0; i < images.length; i++) {
            PDPage page = (PDPage) pdf.getDocumentCatalog().getAllPages()
                    .get(i);
            BufferedImage image;
            try {
//              image = page.convertToImage(BufferedImage.TYPE_INT_RGB, 288); //works
                image = page.convertToImage(BufferedImage.TYPE_INT_RGB, 300); // does not work
                images[i] = image;
            } catch (IOException e) {
                e.printStackTrace();
            }
        }

        FileOutputStream fos = new FileOutputStream("a.tiff");
        RandomAccessOutputStream rout = new FileCacheRandomAccessOutputStream(
                fos);
        ImageParam.ImageParamBuilder builder = ImageParam.getBuilder();
        ImageParam[] param = new ImageParam[1];
        TIFFOptions tiffOptions = new TIFFOptions();
        tiffOptions.setTiffCompression(Compression.CCITTFAX4);
        builder.imageOptions(tiffOptions);
        builder.colorType(ImageColorType.BILEVEL).ditherMatrix(DitherMatrix.getBayer8x8Diag()).applyDither(true).ditherMethod(DitherMethod.BAYER);
        param[0] = builder.build();
        TIFFTweaker.writeMultipageTIFF(rout, param, images);
        rout.close();
        fos.close();
    }
}

For ghostscript, I used command line directly with the same parameters provided by the OP. The screenshots for the first page of the resulted TIFF images are showing below:

The lefthand side shows the output of "ghostscript" and the righthand side the output of "icafe". It can be seen, at least in this case, the output from "icafe" is better than the output from "ghostscript".

Using CCITTFAX4 compression, the file size from "ghostscript" is 2.22M and the file size from "icafe" is 2.08M. Both are not so good given the fact dither is used while creating the black and white output. In fact, a different compression algorithm will create way smaller file size. For example, using LZW, the same output from "icafe" is only 634K and if using DEFLATE compression the output file size went down to 582K.

Solution 2

Here's some code to save in a multipage tiff which I use with PDFBox. It requires the TIFFUtil class from PDFBox (it isn't public, so you have to make a copy).

void saveAsMultipageTIFF(ArrayList<BufferedImage> bimTab, String filename, int dpi) throws IOException
{
    Iterator<ImageWriter> writers = ImageIO.getImageWritersByFormatName("tiff");
    ImageWriter imageWriter = writers.next();

    ImageOutputStream ios = ImageIO.createImageOutputStream(new File(filename));
    imageWriter.setOutput(ios);
    imageWriter.prepareWriteSequence(null);
    for (BufferedImage image : bimTab)
    {
        ImageWriteParam param = imageWriter.getDefaultWriteParam();
        IIOMetadata metadata = imageWriter.getDefaultImageMetadata(new ImageTypeSpecifier(image), param);
        param.setCompressionMode(ImageWriteParam.MODE_EXPLICIT);
        TIFFUtil.setCompressionType(param, image);
        TIFFUtil.updateMetadata(metadata, image, dpi);
        imageWriter.writeToSequence(new IIOImage(image, null, metadata), param);
    }
    imageWriter.endWriteSequence();
    imageWriter.dispose();
    ios.flush();
    ios.close();
}

I experimented on this for myself some time ago by using this code: https://www.java.net/node/670205 (I used solution 2)

However...

If you create an array with lots of images, your memory consumption really goes up. So it would probably be better to render an image, then add it to the tiff file, then render the next page and lose the reference of the previous one so that the gc can get the space if needed.

Solution 3

Since some dependencies used by solutions for this problem looks not maintained. I got a solution by using latest version (2.0.16) pdfbox:

ByteArrayOutputStream imageBaos = new ByteArrayOutputStream();
ImageOutputStream output = ImageIO.createImageOutputStream(imageBaos);
ImageWriter writer = ImageIO.getImageWritersByFormatName("TIFF").next();

try (final PDDocument document = PDDocument.load(new File("/tmp/tmp.pdf"))) {

            PDFRenderer pdfRenderer = new PDFRenderer(document);

            int pageCount = document.getNumberOfPages();

            BufferedImage[] images = new BufferedImage[pageCount];
            // ByteArrayOutputStream[] baosArray = new ByteArrayOutputStream[pageCount];

            writer.setOutput(output);

            ImageWriteParam params = writer.getDefaultWriteParam();

            params.setCompressionMode(ImageWriteParam.MODE_EXPLICIT);

            // Compression: None, PackBits, ZLib, Deflate, LZW, JPEG and CCITT
            // variants allowed
            params.setCompressionType("Deflate");

            writer.prepareWriteSequence(null);

            for (int page = 0; page < pageCount; page++) {
                BufferedImage image = pdfRenderer.renderImageWithDPI(page, DPI, ImageType.RGB);
                images[page] = image;
                IIOMetadata metadata = writer.getDefaultImageMetadata(new ImageTypeSpecifier(image), params);
                writer.writeToSequence(new IIOImage(image, null, metadata), params);
                // ImageIO.write(image, "tiff", baosArray[page]);
            }

            System.out.println("imageBaos size: " + imageBaos.size());
            // Finished write to output

            writer.endWriteSequence();

            document.close();
        } catch (IOException e) {
            e.printStackTrace();
            throw new Exception(e);
        } finally {
            // avoid memory leaks
            writer.dispose();
        }

Then you may using imageBaos write to your local file. But if you want to pass your image to ByteArrayOutputStream and return to privious method like me. Then we need other steps.

After processing is done, the image bytes would be available in the ImageOutputStream output object. We need to position the offset to the beginning of the output object and then read the butes to write to new ByteArrayOutputStream, a concise way like this:

ByteArrayOutputStream bos = new ByteArrayOutputStream();
long counter = 0; 
        while (true) {
            try {
                bos.write(ios.readByte());
                counter++;
            } catch (EOFException e) {
                System.out.println("End of Image Stream");
                break;
            } catch (IOException e) {
                System.out.println("Error processing the Image Stream");
                break;
            }
        }
return bos

Or you can just ImageOutputStream.flush() at end to get your imageBaos Byte then return.

Solution 4

Refer to my github code for an implementation with PDFBox.

View more solutions

Comments

[Raphael Roth]

thanks for your suggestion. But I would prefer to have a solution where I don't need to copy/change private clases from a library

[Uncle Iroh]

Am I the only one that thinks the output on the left is better that the output on the right? In particular the C in PRINCE at the top.