Page wise image segregation

[Deepak-Zais]

in image extraction add the page wise segregation, that helps find the images more quickly and can be mapped easily.

[hnc-jglee]

Hi @Deepak-Zais, thanks for the suggestion!

Let me start by clarifying the current behavior:

Current behavior

Extracted images are saved into a single flat directory {pdf_basename}_images/ with a global index — imageFile1.png, imageFile2.png, ... The page number is not encoded into the directory layout or the filename.

Relevant code:

• ImagesUtils.java
• MarkdownSyntax.java#L28

Page mapping is already available via JSON output

If you run with --format json, each image entry already includes a 1-indexed page number field along with the source path, so you can map images to pages today:

{
  "type": "image",
  "id": 2,
  "page number": 1,
  "bounding box": [1507.354, 45.472, 1675.946, 214.064],
  "source": "mydoc_images/imageFile1.png"
}

So a small post-processing script can move/rename images into per-page folders, or you can use the JSON itself as an index.

About your request

A native option that segregates images on disk by page (e.g., page_1/imageFile1.png or page1_imageFile1.png) does not exist today. Could you share a bit more about your use case — specifically, why the on-disk layout itself needs to be page-segregated (e.g., a workflow that doesn't consume the JSON, per-page OCR, ML dataset preparation, etc.)? That context would help us evaluate the priority.