是否可以使用 iText 获取现有 PDF 文档的元素？

Question

I want to read and get elements of an existing PDF document using iText API. Example: a document contain a PDF table, I want to get that table when reading document.

Answer 1

Directly and easily, no.

If you're willing to put in work, it depends.

If you're willing to put in a lot of work, yes.

Allow me to elaborate. There are 2 flavors of PDF specification. Tagged and untagged PDF. When a PDF is tagged, it means that all the structure information is preserved. Every character belongs to a line, every line belongs to a paragraph, and tables, lists (and other structure elements) know which lines and paragraphs are contained within them.

If you have an untagged PDF, it contains only the instructions needed for rendering the document. You can imagine this as

go to position 50, 50
set the font to Arial Unicode
set the font size to 12
draw the character 'H'

This is where the solution depends on the amount of work. If your PDF is tagged, you can use iText to extract the tagging information, and this allows you to rebuild a structural concept of PdfTable. (you can also use IEventListener to find the font that was used, the font size, etc)

If your PDF is untagged, you can attempt to find the structure in the rendering instructions.

This is a hard problem. The topic of research even. Two main approaches seem to exist in current research:

• Rule based (characters are considered part of the same line if their distance is smaller than a given epsilon, and their y-position is roughly the same within a given margin, etc)
• Neural network ("render the PDF" and treat the image as the input for an image classification network)