Embed Base64-Encoded Images Inline

[HiromuHota]

Description of the problems or issues

Is your pull request related to a problem? Please describe.

See #88

Does your pull request fix any issue.

Close #88

Description of the proposed changes

Embed base64-encoded images inline. Support starting with JPEG and BMP.

Test plan

Apply pdftotree to pdfs and see if JPEG and BMP images are extracted.

Checklist

• [ ] I have updated the documentation accordingly.
• [x] I have added tests to cover my changes.
• [x] All new and existing tests passed.
• [x] I have updated the CHANGELOG.md accordingly.

[codecov-io]

Codecov Report

Merging #99 into master will increase coverage by 0.42%. The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master      #99      +/-   ##
==========================================
+ Coverage   66.40%   66.83%   +0.42%     
==========================================
  Files          23       24       +1     
  Lines        2566     2593      +27     
==========================================
+ Hits         1704     1733      +29     
+ Misses        862      860       -2     

Flag	Coverage Δ
#unittests	66.83% <100.00%> (+0.42%)	:arrow_up:

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
pdftotree/TreeExtract.py	89.53% <100.00%> (+0.79%)	:arrow_up:
pdftotree/ml/features.py	64.58% <100.00%> (ø)
pdftotree/utils/bbox_utils.py	21.91% <100.00%> (+4.52%)	:arrow_up:
pdftotree/utils/pdf/node.py	45.39% <100.00%> (+0.39%)	:arrow_up:
pdftotree/utils/pdf/pdf_parsers.py	92.61% <100.00%> (+0.11%)	:arrow_up:
tests/test_figures.py	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update d4e5dc8...1cfa226. Read the comment docs.

[lukehsiao]

What happens today for a PNG?

Adobe goes with the alternative you mention in the issue (saving it as a file, and pointing to that instead), which seems to work fairly well.

[lukehsiao]

Ah, I see now that it looks like you can just specify the MIME type, so we should be able to support a wide variety of image formats this way, right? Basically, I want to check that this approach will be support a wide variety of images.

[HiromuHota]

Ah, I see now that it looks like you can just specify the MIME type, so we should be able to support a wide variety of image formats this way, right? Basically, I want to check that this approach will be support a wide variety of images.

Yes, this approach can support a wide variety of image types. https://developer.mozilla.org/en-US/docs/Web/HTTP/Basics_of_HTTP/Data_URIs In fact, JPEG and BMP are just an initial set of supported types. We will add PNG and other types too when we have such an example PDF containing different types of image.

I like the fact that hOCR with embedded images is more portable than hOCR with multiple images. Please let me know your thoughts and experiences.

[lukehsiao]

I like the fact that hOCR with embedded images is more portable than hOCR with multiple images. Please let me know your thoughts and experiences.

I don't know about portable vs not portable, but I think embedding them seems just fine. I assume I could download the embedded image as a file anyways, right?

[HiromuHota]

@lukehsiao Yes, you can open an exported hocr and save an embedded image as a separate file. In the future, pdftotree might have options like --embed-image <0|1> (Default: 1) as in pdf2htmlEX. https://github.com/coolwanglu/pdf2htmlEX/wiki/Command-Line-Options With this, you can control whether images are embedded or saved as a separate file.

[lukehsiao]

Sounds great. Feel free to merge whenever you'd like.

[HiromuHota]

For future reference, this PR relies on pdfminer.six for image type detection, which cannot detect an image type sometimes. This PR embeds images only when their image types are detected and does not embed the image of unknown type. To detect an image type on broader cases, we have to

• Enhance and contribute to pdfminer.six, or
• Use a different library or implement it on our own.