question: jpeg compression?

[dmaziuk]

Is there a simple way to use jpeg compression instead of deflate=True, like tif2pdf's -j? -- zlib compression is useless in practice as half the readers can't read the resulting PDFs, and no compression is useless too if your PDFs are large enough.

[JorjMcKie]

Please have a look at Document method rewrite_images. It represents pretty much all we have to offer here.

[dmaziuk]

Got it, thanks. It is better than having to manually add each image as a compressed JPEG, so there's that.

[robinwatts]

"half the readers can't read the resulting PDFs" - I'm sorry?

Flate compression has been in the spec since v1.1, I believe, over 30 years ago.

EVERY PDF reader can cope with zlib compressed data/images, or they aren't PDF readers.

[dmaziuk]

After actually trying: running it with dpi_threshold = 101, dpi_target = 100, quality = 75 doesn't do anything: I get the same 3GB PDF that I get w/o rewrite_images() and compression. Looks like a bug, at least as documented, to me.

For comparison tiff2pdf -j -x 100 -y 100 -q 75 generates a ~200K PDF.

FWIW I get about the same size PDF by adding each page as a 75-quality jpeg, as I originally thought. (And it opens in chrominum, too.)

[JorjMcKie]

dpi_threshold and dpi_target should differ quite a lot - say 20 or more.

[dmaziuk]

no change using dpi_threshold = 150, dpi_target = 100

This is when adding pages as PNGs (converted from TIFF by PIL). I wonder if it doesn't actually do anything to "lossless" images in the first place, though my reading of TFM is that it should?

Edit: hmm... maybe I should try reading tif into a pixmap and inserting that in a page ...

[dmaziuk]

... and that doesn't work because tiff support appears to be... lacking, compared to libtiff or PIL. So the only thing that works is inserting pages as JPEGs pre-compressed to the desired quality.

Interestingly enough, calling rewrite_images() on top results in a larger pdf: 315MB without, 340MB with.