[Parsing] Images nested in Form XObjects are not detected
Author: Tsan1024Created Aug 24, 2026Updated Aug 28, 2026
Description
Raster images nested inside PDF Form XObjects are not detected by LiteParse.
Images attached directly to a page are discovered correctly, but images contained in a Form XObject are omitted, even when there is only one level of nesting.
This can cause:
- Incorrect per-page image counts and image coverage.
- Missing image references in Markdown output.
- Missing images when image extraction is enabled.
- Incorrect image-related complexity or OCR decisions.
Expected behavior
LiteParse should recursively discover image objects contained in Form XObjects and return their bounds in page coordinates.
Actual behavior
Only top-level page image objects are discovered. Image objects inside Form XObjects are not returned.
Source: run-llama/liteparse