从PDF创建图像（PNG或JPEG）以及图像中文本的HTML图像映射？

我正在记录我维护的系统。此文档包含我在TeX/TikZ中创建的一个图表，该图表将呈现为PDF文件。然后我将PDF文件转换为图像文件（通过imagemagick转换为PNG），并将其包含在HTML文档中。效果很好

现在，我想为该图像创建一个，这样我就可以添加超链接/鼠标盖/等等。这是一个我希望根据系统中的更改定期更新的图像，所以如果可能的话，我想自动化这个过程

当PDF文件渲染为PNG时，是否有方法使用软件库或工具自动创建PDF文件中各种文本内容的图像映射？

以下是我创建的一个示例：

在本例中，我希望通过在PDF中定位各种文本字符串的边界框，将其转换为超链接：

• 控制器

• 执行器

• 传感器

• A

• B

• C

• D

• u

• y

• F（s）

• G（s）

• H（s）

（它们都是PDF文件中的文本内容；我可以在Acrobat Reader中选择其中任何一个的文本，然后复制并粘贴到我的文本编辑器中。）

---

有办法做到这一点吗？

嗯。我找到了ApachePDFBox库，其中包含一个名为的示例，它可以打印信息，但我不确定如何解释它，每个glyph只有一个位置

> java -jar print_text_locations.jar blockdiagram_example.pdf
String[37.864998,13.939003 fs=4.9813 xscale=4.9813 height=2.49065 space=2.4906502 width=5.1197815]+
String[59.185997,13.662003 fs=9.9626 xscale=9.9626 height=6.1668496 space=2.769603 width=6.6450577]A
String[130.229,13.662003 fs=9.9626 xscale=9.9626 height=6.1668496 space=2.769603 width=6.64505]B
String[198.783,13.498001 fs=9.9626 xscale=9.9626 height=6.1668496 space=2.769603 width=7.192993]C
String[86.827,21.278 fs=11.9552 xscale=11.9552 height=5.9776 space=5.9776006 width=9.699257]H
String[97.449005,21.278 fs=11.9552 xscale=11.9552 height=5.983578 space=5.9776006 width=4.552536](
String[102.00201,21.278 fs=11.9552 xscale=11.9552 height=5.9776 space=5.9776006 width=5.5137405]s
String[107.51601,21.278 fs=11.9552 xscale=11.9552 height=5.983578 space=5.9776006 width=4.552536])
String[156.35,21.278 fs=11.9552 xscale=11.9552 height=5.9776 space=5.9776006 width=9.234192]G
String[165.58301,21.278 fs=11.9552 xscale=11.9552 height=5.983578 space=5.9776006 width=4.552536](
String[170.136,21.278 fs=11.9552 xscale=11.9552 height=5.9776 space=5.9776006 width=5.513733]s
String[175.65,21.278 fs=11.9552 xscale=11.9552 height=5.983578 space=5.9776006 width=4.552536])
String[12.797,29.332 fs=9.9626 xscale=9.9626 height=4.9813 space=4.9813004 width=5.7035875]u
String[38.711,27.432999 fs=4.9813 xscale=4.9813 height=3.4022279 space=2.4906502 width=5.39624]?
String[214.641,29.332 fs=9.9626 xscale=9.9626 height=4.9813 space=4.9813004 width=4.884659]y
String[85.109,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.4869003]c
String[88.5959,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774338]o
String[92.473335,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774338]n
String[96.35077,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=1.9387131]t
String[98.28948,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=2.3222733]r
String[100.611755,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774338]o
String[104.48919,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=1.5481873]l
String[106.03738,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=1.5481873]l
String[107.58556,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774338]e
String[111.463,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=2.3222733]r
String[155.67801,41.046 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774261]a
String[159.55544,41.046 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.4868927]c
String[163.04233,41.046 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=1.9387207]t
String[164.98105,41.046 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774261]u
String[168.85847,41.046 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774261]a
String[172.7359,41.046 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=1.9387207]t
String[174.67462,41.046 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774261]o
String[178.55205,41.046 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=2.322281]r
String[58.912003,65.483 fs=9.9626 xscale=9.9626 height=6.1668496 space=2.769603 width=7.192993]D
String[87.536,73.099 fs=11.9552 xscale=11.9552 height=5.9776 space=5.9776006 width=7.577202]F
String[96.740005,73.099 fs=11.9552 xscale=11.9552 height=5.983578 space=5.9776006 width=4.552536](
String[101.29201,73.099 fs=11.9552 xscale=11.9552 height=5.9776 space=5.9776006 width=5.5137405]s
String[106.80601,73.099 fs=11.9552 xscale=11.9552 height=5.983578 space=5.9776006 width=4.5525436])
String[88.983,91.978004 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.4869003]s
String[92.4699,91.978004 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774338]e
String[96.347336,91.978004 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774338]n
String[100.22477,91.978004 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.4869003]s
String[103.71167,91.978004 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774338]o
String[107.5891,91.978004 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=2.3222733]r

但是，我做了一个小的更改，看起来每个文本项都会调用

writeString

方法，我想我可以找到每个字符串的整体边框：

/**
 * Override the default functionality of PDFTextStripper.
 */
@Override
protected void writeString(String string, List<TextPosition> textPositions) throws IOException
{
    System.out.println("text string: "+string);
    for (TextPosition text : textPositions)
    {
        System.out.println( "String[" + text.getXDirAdj() + "," +
                text.getYDirAdj() + " fs=" + text.getFontSize() + " xscale=" +
                text.getXScale() + " height=" + text.getHeightDir() + " space=" +
                text.getWidthOfSpace() + " width=" +
                text.getWidthDirAdj() + "]" + text.getUnicode() );
    }
}

---

嗯，我找到了ApachePDFBox库，它包含一个名为的示例，它可以打印信息，但我不确定如何解释它，每个glyph有一个位置

> java -jar print_text_locations.jar blockdiagram_example.pdf
String[37.864998,13.939003 fs=4.9813 xscale=4.9813 height=2.49065 space=2.4906502 width=5.1197815]+
String[59.185997,13.662003 fs=9.9626 xscale=9.9626 height=6.1668496 space=2.769603 width=6.6450577]A
String[130.229,13.662003 fs=9.9626 xscale=9.9626 height=6.1668496 space=2.769603 width=6.64505]B
String[198.783,13.498001 fs=9.9626 xscale=9.9626 height=6.1668496 space=2.769603 width=7.192993]C
String[86.827,21.278 fs=11.9552 xscale=11.9552 height=5.9776 space=5.9776006 width=9.699257]H
String[97.449005,21.278 fs=11.9552 xscale=11.9552 height=5.983578 space=5.9776006 width=4.552536](
String[102.00201,21.278 fs=11.9552 xscale=11.9552 height=5.9776 space=5.9776006 width=5.5137405]s
String[107.51601,21.278 fs=11.9552 xscale=11.9552 height=5.983578 space=5.9776006 width=4.552536])
String[156.35,21.278 fs=11.9552 xscale=11.9552 height=5.9776 space=5.9776006 width=9.234192]G
String[165.58301,21.278 fs=11.9552 xscale=11.9552 height=5.983578 space=5.9776006 width=4.552536](
String[170.136,21.278 fs=11.9552 xscale=11.9552 height=5.9776 space=5.9776006 width=5.513733]s
String[175.65,21.278 fs=11.9552 xscale=11.9552 height=5.983578 space=5.9776006 width=4.552536])
String[12.797,29.332 fs=9.9626 xscale=9.9626 height=4.9813 space=4.9813004 width=5.7035875]u
String[38.711,27.432999 fs=4.9813 xscale=4.9813 height=3.4022279 space=2.4906502 width=5.39624]?
String[214.641,29.332 fs=9.9626 xscale=9.9626 height=4.9813 space=4.9813004 width=4.884659]y
String[85.109,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.4869003]c
String[88.5959,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774338]o
String[92.473335,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774338]n
String[96.35077,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=1.9387131]t
String[98.28948,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=2.3222733]r
String[100.611755,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774338]o
String[104.48919,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=1.5481873]l
String[106.03738,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=1.5481873]l
String[107.58556,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774338]e
String[111.463,41.419 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=2.3222733]r
String[155.67801,41.046 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774261]a
String[159.55544,41.046 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.4868927]c
String[163.04233,41.046 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=1.9387207]t
String[164.98105,41.046 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774261]u
String[168.85847,41.046 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774261]a
String[172.7359,41.046 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=1.9387207]t
String[174.67462,41.046 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774261]o
String[178.55205,41.046 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=2.322281]r
String[58.912003,65.483 fs=9.9626 xscale=9.9626 height=6.1668496 space=2.769603 width=7.192993]D
String[87.536,73.099 fs=11.9552 xscale=11.9552 height=5.9776 space=5.9776006 width=7.577202]F
String[96.740005,73.099 fs=11.9552 xscale=11.9552 height=5.983578 space=5.9776006 width=4.552536](
String[101.29201,73.099 fs=11.9552 xscale=11.9552 height=5.9776 space=5.9776006 width=5.5137405]s
String[106.80601,73.099 fs=11.9552 xscale=11.9552 height=5.983578 space=5.9776006 width=4.5525436])
String[88.983,91.978004 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.4869003]s
String[92.4699,91.978004 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774338]e
String[96.347336,91.978004 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774338]n
String[100.22477,91.978004 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.4869003]s
String[103.71167,91.978004 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=3.8774338]o
String[107.5891,91.978004 fs=6.9738 xscale=6.9738 height=4.3167825 space=1.9387167 width=2.3222733]r

但是，我做了一个小的更改，看起来每个文本项都会调用

writeString

方法，我想我可以找到每个字符串的整体边框：

/**
 * Override the default functionality of PDFTextStripper.
 */
@Override
protected void writeString(String string, List<TextPosition> textPositions) throws IOException
{
    System.out.println("text string: "+string);
    for (TextPosition text : textPositions)
    {
        System.out.println( "String[" + text.getXDirAdj() + "," +
                text.getYDirAdj() + " fs=" + text.getFontSize() + " xscale=" +
                text.getXScale() + " height=" + text.getHeightDir() + " space=" +
                text.getWidthOfSpace() + " width=" +
                text.getWidthDirAdj() + "]" + text.getUnicode() );
    }
}

---

我能够将以下Python解决方案组合在一起，作为一个起点。它将pdf转换为png并输出相应的图像地图标记

它将输出dpi作为可选参数（默认值200），以便将边界框从默认的pdf dpi 72正确缩放到png：

从pdf2image导入从路径转换
从pdfminer.converter导入PDFPageAggregator
从pdfminer.layout导入LAParams，LTTextBox
从pdfminer.pdfinterp导入pdfpageexplorer
从pdfminer.pdfinterp导入PDFResourceManager
从pdfminer.pdfpage导入pdfpage
从标签导入单据，缩进
导入argparse
导入操作系统
def转换坐标（lobj，mb）：
#将LTTextBox边界框转换为图像贴图区域边界框。
#
#每个LTTextBox的边界框指定为：
#
#x0：从页面左侧到框左边缘的距离
#y0：从页面底部到框的下边缘的距离
#x1：从页面左侧到方框右边缘的距离
#y1：从页面底部到框的上边缘的距离
#
#所以y坐标从图像的底部开始。但是有了图像地图
#面积，y坐标从图像顶部开始，所以这里我们减去
#边界框的y轴值来自总高度。
返回[lobj.x0，mb[3]-lobj.y1，lobj.x1，mb[3]-lobj.y0]
def get_图像映射（d）：
doc，tag，text=doc（）.tagtext（）
带标签（“地图”，name=“地图”）：
对于d.项（）中的k，v：
doc.stag（“area”，shape=“rect”，coords=”，“.join（v），href=”，alt=k）
返回缩进（doc.getvalue（））
def get_bboxes（pdf，dpi）：
fp=开放（pdf，“rb”）
rsrcmgr=PDFResourceManager（）
device=PDFPageAggregator（rsrcmgr，laparams=laparams（））
解释器=PDFPAGE解释器（rsrcmgr，设备）
页面=列表（PDFPage.get_页面（fp））[0]
解释器。处理页面（第页）
布局=设备。获取_结果（）
#PDFminer基于72的dpi报告边界框。我找不到办法
#为了改变这一点，我用dpi/72来缩放每个坐标
刻度=dpi/72.0
返回{
lobj.get_text（）.strip（）：[
转换坐标中x的str（int（x*scale））（lobj，page.mediabox）
]
用于布局中的lobj
如果isinstance（lobj，LTTextBox）
}
def main（）：
parser=argparse.ArgumentParser（）
parser.add_参数（“pdf”）
add_参数（“--dpi”，type=int，default=200）
args=parser.parse_args（）
页面=列表（从路径（args.pdf，args.dpi）转换）[0]
save（f“{os.path.splitext（args.pdf）[0]}.png”，“png”）
打印（get_imagemap（get_bboxes（args.pdf，args.dpi）））
如果名称=“\uuuuu main\uuuuuuuu”：
main（）

示例结果：

---

我能够将以下Python解决方案组合在一起，作为一个起点。它将pdf转换为png并输出相应的图像地图标记

它将输出dpi作为可选参数（默认值200），以便将边界框从默认的pdf dpi 72正确缩放到png：

从pdf2image导入从路径转换
从pdfminer.converter导入PDFPageAggregator
从pdfminer.layout导入LAParams，LTTextBox
从pdfminer.pdfinterp导入pdfpageexplorer
从pdfminer.pdfinterp导入PDFResourceManager
从pdfminer.pdfpage导入pdfpage
从标签导入单据，缩进
导入argparse
导入操作系统
def转换坐标（lobj，mb）：
#将LTTextBox边界框转换为图像贴图区域边界框。
#
#每个LTTextBox的边界框指定为：
#
#x0：从页面左侧到页面左边缘的距离