PDF - Only OCR Images and Background Images

[saschacontes]

Good day Community,

to optimize performance, plain text from PDFs should not undergo OCR, but images/pictures and background imagesshould undergo external vlm OCR and the resulting text/descriptuon/annotaion should be inserted at the correct location in the markdown export.

I identified that these images/pictures are located in two different places:
pictures[no].image.uri
pages[no].image.uri

Thanks and Best Regards,
Sascha

Edit: Correction, the image/s in question are not showing in the Docling document structure.

[nunolosada]

I have a similar issue, but I see some images.
I want to see independent images present in the pdf, but I can't find them.
I can get each of the pages image in the
"pages": {
"1": {
"size": {
"width": 595,
"height": 841
},
"image": {
"mimetype": "image/png",
"dpi": 144,
"size": {
"width": 1190,
"height": 1682
},
"uri": "data:image/png;base64,{{page 1 image in base64 format}}
},
"page_no": 1
},
"2": {
"size": {
"width": 595,
"height": 841
},
"image": {
"mimetype": "image/png",
"dpi": 144,
"size": {
"width": 1190,
"height": 1682
},
"uri": "data:image/png;base64,{{page 2 image in base64 format}}
},
"page_no": 2
},
...

[saschacontes]

Interestingly docling-parse finds them all.