如何减小 pdf 中 png 图像的大小（在 pdf 中压缩 png）

Question

我想通过用低分辨率图像替换高分辨率图像来减小 pdf 文件的大小。 要完成这个问题，我必须：

1. 从 pdf 中提取图像（流）
2. 压缩图像
3. 用压缩图像替换 pdf 中的图像（流）

当我提取 png 图像并替换它们时，透明背景变为黑色背景。 我从 pdf 中提取图像以找出原因。 pdf 使用 stream 保存 png 有一些非常奇怪的东西。 因此，如果我尝试从 pdf 中提取 png 图像，我将得到两个不同的图像：一个 8 位彩色图像和一个 24 位彩色图像。

...
1 0 obj
<</Type/XObject/Subtype/Image/Width 1920/Height 1035/Length 24720/ColorSpace/DeviceGray/BitsPerComponent 8/Filter/FlateDecode>>stream
...
endstream
endobj
2 0 obj
<</Type/XObject/Subtype/Image/Width 1920/Height 1035/SMask 1 0 R/Length 47751/ColorSpace[/CalRGB<</Gamma[2.2 2.2 2.2]/Matrix[0.41239 0.21264 0.01933 0.35758 0.71517 0.11919 0.18045 0.07218 0.9504]/WhitePoint[0.95043 1 1.09]>>]/Intent/Perceptual/BitsPerComponent 8/Filter/FlateDecode>>stream
...
endstream
...

原始图像（透明背景的 32 位彩色图像）：

8 位彩色图像：

24 位彩色图像：

<dependency>
    <groupId>com.itextpdf</groupId>
    <artifactId>itextpdf</artifactId>
    <version>5.5.12</version>
</dependency>
<dependency>
    <groupId>org.apache.pdfbox</groupId>
    <artifactId>pdfbox</artifactId>
    <version>2.0.16</version>
</dependency>

ImageExtractor将帮助您从 Pdf 文件中提取图像。

public class ImageExtractor {

    private static final Logger log = LoggerFactory.getLogger(ImageExtractor.class);

    public void extract(File pdf, File imageDir) throws IOException {
        if(!imageDir.exists()) {
            imageDir.mkdirs();
        }
        PDDocument document = PDDocument.load(pdf);
        PDPageTree list = document.getPages();
        System.out.println("PDPageTree#count: " + list.getCount());
        int pageIndex = 1;
        for (PDPage page : list) {
            PDResources pdResources = page.getResources();
            System.out.println(pdResources.toString());
            for (COSName c : pdResources.getXObjectNames()) {
                System.out.println("PDResources[" + pageIndex + "]#COSName: " + c.getName());
                PDXObject o = pdResources.getXObject(c);
                System.out.println("PDResources[" + pageIndex + "]#PDXObject: " + o.toString());
                // https://github.com/mkl-public/testarea-itext5/blob/master/src/test/java/mkl/testarea/itext5/extract/ImageExtraction.java
                if (o instanceof PDImageXObject) {
                    PDImageXObject img = (PDImageXObject) o;
                    File file = new File(imageDir, pageIndex + "-" + System.nanoTime() + "." + img.getSuffix());
                    ImageIO.write(((PDImageXObject)o).getImage(), img.getSuffix(), file);
                }
            }
            pageIndex ++;
        }
        log.info("Images have been extracted successfully! Check your images folder.");
    }
}

ReplaceHightResolutionImage是我用来减小 pdf 大小的代码。

package io.gitlab.donespeak.tutorial.pdf.reducesize.itext;

import com.itextpdf.text.DocumentException;
import com.itextpdf.text.pdf.PRStream;
import com.itextpdf.text.pdf.PdfName;
import com.itextpdf.text.pdf.PdfNumber;
import com.itextpdf.text.pdf.PdfObject;
import com.itextpdf.text.pdf.PdfReader;
import com.itextpdf.text.pdf.PdfStamper;
import com.itextpdf.text.pdf.PdfStream;
import com.itextpdf.text.pdf.parser.PdfImageObject;
import io.gitlab.donespeak.tutorial.pdf.reducesize.imagecompress.ImageCompressor;
import io.gitlab.donespeak.tutorial.pdf.reducesize.imagecompress.SimpleCompress;

import javax.imageio.ImageIO;
import java.awt.image.BufferedImage;
import java.io.ByteArrayOutputStream;
import java.io.File;
import java.io.FileInputStream;
import java.io.FileOutputStream;
import java.io.IOException;

public class ReplaceHightResolutionImage {

    private ImageCompressor compressor;
    private double quality;
    private double scale;

    public ReplaceHightResolutionImage(double quality, double scale) {
        this.compressor = new SimpleCompress();
        this.quality = quality;
        this.scale = scale;
    }

    public ReplaceHightResolutionImage(double quality, double scale, ImageCompressor compressor) {
        this.compressor = compressor;
        this.quality = quality;
        this.scale = scale;
    }

    public void replace(File pdf, File output) throws IOException, DocumentException {
        PdfReader reader = new PdfReader(new FileInputStream(pdf));
        int n = reader.getXrefSize();
        PdfObject object;
        PRStream stream;

        for (int i = 0; i < n; i++) {

            object = reader.getPdfObject(i);
            stream = findImageStream(object);
            if (stream == null) {
                continue;
            }
            PdfImageObject pdfImageObject = new PdfImageObject(stream);
            BufferedImage bi = pdfImageObject.getBufferedImage();
            if (bi == null) {
                continue;
            }
            System.out.println("PdfReader#Xref: " + i + "," + pdfImageObject.getFileType());
            BufferedImage resultImage = compressor.compress(bi, pdfImageObject.getFileType(), quality, scale);
            replaceImage(stream, resultImage);
        }

        PdfStamper stamper = new PdfStamper(reader, new FileOutputStream(output));
        // furtherCompress(reader, stamper);
        stamper.close();
    }

    private void furtherCompress(PdfReader reader, PdfStamper stamper) throws DocumentException {
        reader.removeFields();
        reader.removeUnusedObjects();
        stamper.setFullCompression();
        stamper.getWriter().setCompressionLevel(PdfStream.DEFAULT_COMPRESSION);
    }

    private PRStream findImageStream(PdfObject object) {
        PRStream stream;
        if (object == null || !object.isStream()) {
            return null;
        }
        stream = (PRStream)object;
        System.out.println(stream.getAsName(PdfName.SUBTYPE));
        if (!PdfName.IMAGE.equals(stream.getAsName(PdfName.SUBTYPE))) {
            // not jpg or png
            return null;
        }
        PdfName pdfName = stream.getAsName(PdfName.FILTER);
        if (!PdfName.DCTDECODE.equals(pdfName) && !PdfName.FLATEDECODE.equals(pdfName)) {
            return null;
        }
        // if (PdfName.DCTDECODE.equals(filter)) {
        //     return PdfImageObject.ImageBytesType.JPG.getFileExtension();
        // } else if (PdfName.JPXDECODE.equals(filter)) {
        //     return PdfImageObject.ImageBytesType.JP2.getFileExtension();
        // } else if (PdfName.FLATEDECODE.equals(filter)) {
        //     return PdfImageObject.ImageBytesType.PNG.getFileExtension();
        // } else if (PdfName.LZWDECODE.equals(filter)) {
        //     return PdfImageObject.ImageBytesType.CCITT.getFileExtension();
        // }
        return stream;
    }

    private void replaceImage(PRStream stream, BufferedImage resultImage) throws IOException {

        ByteArrayOutputStream imgBytes = new ByteArrayOutputStream();
        ImageIO.write(resultImage, "JPG", imgBytes);

        stream.clear();
        stream.setData(imgBytes.toByteArray(), false, PRStream.NO_COMPRESSION);
        stream.put(PdfName.TYPE, PdfName.XOBJECT);
        stream.put(PdfName.SUBTYPE, PdfName.IMAGE);
        stream.put(PdfName.FILTER, PdfName.DCTDECODE);
        stream.put(PdfName.WIDTH, new PdfNumber(resultImage.getWidth()));
        stream.put(PdfName.HEIGHT, new PdfNumber(resultImage.getHeight()));
        stream.put(PdfName.BITSPERCOMPONENT, new PdfNumber(8));
        stream.put(PdfName.COLORSPACE, PdfName.DEVICERGB);
    }
}

package io.gitlab.donespeak.tutorial.pdf.reducesize.itext;

public class ThumbnailatorCompressor implements ImageCompressor {

    @Override
    public BufferedImage compress(BufferedImage image, String imageFormat, double quality, double scale) throws IOException {
        System.out.println("ThumbnailatorCompressor#type: " + image.getType());
        // int imageType = "png".equalsIgnoreCase(imageFormat)? BufferedImage.TYPE_INT_ARGB: image.getType();
        BufferedImage thumbnail = Thumbnails.of(image)
            .imageType(image.getType())
            .scale(scale)
            .outputQuality(quality)
            // .outputFormat(imageFormat)
            .useOriginalFormat()
            .asBufferedImage();

        return thumbnail;
    }
}

• 马.pdf
• 马.png

public class ReplaceHightResolutionImageTest {

    @Test
    public void reduceWithThumbnailatorCompressor() throws IOException, DocumentException {
        double quality = 1d;
        double scale = 0.6d;
        File pdf = new File("pdf/asset/horse.pdf");
        File output = new File("pdf/target/output", "replaced-" + quality + "-" + scale);
        ReplaceHightResolutionImage replacer = new ReplaceHightResolutionImage(quality, scale, new SimpleCompress());
        replacer.replace(pdf, output);
    }
}

Answer 1

这是一个可行但不够好的答案。 它可以很好地压缩 jpg 和 png。 唯一的缺点是，如果您在许多页面中重复使用图像，它将把每个图像引用作为单独的 stream 并生成一个新的 stream 来代替图像引用，这可能会导致更大的文件大小。

1 0 obj
<</Type/XObject/Subtype/Image/Width 1002/Height 564/Filter/DCTDecode/ColorSpace/DeviceRGB/BitsPerComponent 8/Length 89149>>stream
...
endstream
endobj
2 0 obj
<</Length 106/Filter/FlateDecode>>stream
x�m�=� ��w�^@|���=� 7�/����8�6��&b0$��
��N!o��L�,?Ck'�����c�h�x0��/(5c*�Y�سEX�o�Uj3�B�ݔ"
endstream
endobj
4 0 obj
<</Type/Page/MediaBox[0 0 595 842]/Resources<</XObject<</img0 1 0 R>>>>/Contents 2 0 R/Parent 3 0 R>>
endobj
5 0 obj
<</Length 106/Filter/FlateDecode>>stream
x�m�=� ��w�^@|���=�image    7�/����8�6��&b0$��
��N!o��L�,?Ck'�����c�h�x0��/(5c*�Y�سEX�o�Uj3�B�ݔ"
endstream
endobj
6 0 obj
<</Type/Page/MediaBox[0 0 595 842]/Resources<</XObject<</img0 1 0 R>>>>/Contents 5 0 R/Parent 3 0 R>>
endobj

package io.gitlab.donespeak.tutorial.pdf.reducesize;

import io.gitlab.donespeak.tutorial.pdf.reducesize.imagecompress.ThumbnailatorCompressor;
import org.apache.pdfbox.cos.COSName;
import org.apache.pdfbox.pdmodel.PDDocument;
import org.apache.pdfbox.pdmodel.PDDocumentCatalog;
import org.apache.pdfbox.pdmodel.PDPage;
import org.apache.pdfbox.pdmodel.PDResources;
import org.apache.pdfbox.pdmodel.graphics.PDXObject;
import org.apache.pdfbox.pdmodel.graphics.image.JPEGFactory;
import org.apache.pdfbox.pdmodel.graphics.image.LosslessFactory;
import org.apache.pdfbox.pdmodel.graphics.image.PDImageXObject;

import javax.imageio.ImageIO;
import java.awt.image.BufferedImage;
import java.io.File;
import java.io.IOException;

public class RemoveAllImageFromPdf {

    public static void extractImages(File input, File imageDir) throws IOException {
        if(imageDir.exists()) {
            imageDir.delete();
        }
        imageDir.mkdirs();
        PDDocument document = PDDocument.load(input);
        int pageIndex = 1;
        PDDocumentCatalog catalog = document.getDocumentCatalog();
        for (PDPage page : catalog.getPages()) {
            PDResources pdResources = page.getResources();
            System.out.println(pdResources.toString());
            for (COSName c : pdResources.getXObjectNames()) {
                System.out.println("PDResources[" + pageIndex + "]#COSName: " + c.getName());
                PDXObject o = pdResources.getXObject(c);
                System.out.println("PDResources[" + pageIndex + "]#PDXObject: " + o.toString());
                // https://github.com/mkl-public/testarea-itext5/blob/master/src/test/java/mkl/testarea/itext5/extract/ImageExtraction.java
                if (o instanceof PDImageXObject) {
                    PDImageXObject img = (PDImageXObject) o;
                    System.out.println(img.getSuffix() + "-" + img.getBitsPerComponent() + "-" + img.getColorSpace());
                    File file = new File(imageDir, pageIndex + "-" + c.getName() + "-" + img.getColorSpace() + "-" + System.nanoTime() + "." + img.getSuffix());
                    ImageIO.write(((PDImageXObject)o).getImage(), img.getSuffix(), file);
                }
            }
            pageIndex ++;
        }
        // document.save(output);
    }

    /**
     *
     * @param input
     * @param output
     * @throws IOException
     */
    public static void compress(File input, File output) throws IOException {
        if(!output.getParentFile().exists()) {
            output.getParentFile().mkdirs();
        }
        ThumbnailatorCompressor compressor = new ThumbnailatorCompressor();
        PDDocument document = PDDocument.load(input);
        int pageIndex = 1;
        PDDocumentCatalog catalog = document.getDocumentCatalog();

        for (PDPage page : catalog.getPages()) {
            PDResources pdResources = page.getResources();
            for (COSName c : pdResources.getXObjectNames()) {
                System.out.println("PDResources[" + pageIndex + "]#COSName: " + c.getName());
                PDXObject o = pdResources.getXObject(c);
                System.out.println("PDResources[" + pageIndex + "]#PDXObject: " + o.toString());
                // https://github.com/mkl-public/testarea-itext5/blob/master/src/test/java/mkl/testarea/itext5/extract/ImageExtraction.java
                if (o instanceof PDImageXObject) {
                    PDImageXObject img = (PDImageXObject) o;
                    BufferedImage bufferedImage = compressor.compress(img.getImage(), img.getSuffix(), 0.8, 0.5);
                    PDImageXObject imgNew = null;
                    System.out.println("img(w, h): (" + img.getWidth() + "," + img.getHeight() + ")");
                    System.out.println("bufferedImage(w, h): (" + bufferedImage.getWidth() + "," + bufferedImage.getHeight() + ")");
                    if("png".equalsIgnoreCase(img.getSuffix())) {
                        imgNew = LosslessFactory.createFromImage(document, bufferedImage);
                    } else {
                        imgNew = JPEGFactory.createFromImage(document, bufferedImage);
                    }
                    pdResources.put(c, imgNew);
                }
            }
            pageIndex ++;
        }
        if(!output.getParentFile().exists()) {
            output.getParentFile().mkdirs();
        }
        document.save(output);
        document.close();
    }
}

通过以下方法直接处理文档中的对象，或许可以解决上述问题。 但我不知道如何以这种方式更换 stream。

new com.itextpdf.text.pdf.PdfReader(new FileInputStream(pdf)).getPdfObject(i);
// or
org.apache.pdfbox.pdmodel.PDDocument.load(pdf).getDocument().getObjects()