Publisher 1.8.3 (Windows) -- PDF output: unexpected results of JPEG compression

[MikeA]

In making a book of photographs to be printed by a print-on-demand company, I found I'd exceeded their 300 MB per PDF limit even before I had all image files placed in the Publisher document. The idea of further JPEG compression was not appealing but I decided (holding my nose the entire time) to try compression settings during export to PDF. The results surprised me.

No compression: 388.5 MB
No compression (2nd time around): 363.4 MB (why the difference?)
Compression setting 100: 211 MB
Compression setting 99: 182.7 MB
Compression setting 97: 139 MB
Compression setting 95: 114 MB
Compression setting 90: 83 MB

I take it this means that the max quality setting (100) still involves re-compression of the images. All along I'd thought 100 means: no compression. Apparently not. So if output file sizes are plaguing you, perhaps the slight JPEG compression will be helpful.

Before starting the tests I ran the original JPEGs through the compression program JPEG-Mini. It made a whopping difference in the files' sizes and without noticeable loss of quality. But using the JPEGs compressed with JPEG-Mini as source files within the Publisher document doesn't seem to have provided a lot of savings during export to PDF. Perhaps JPEG-Mini's major advantage is a "web thing" and not a "print thing".

 

[lacerto]

(...)

[MikeA]

I appreciate your taking the time to do all that. I made a note of most of the file sizes I obtained after doing some more of this (compatibility setting for all exports from Publisher: PDF 1.4):

Files marked "[C]" were done with JPEG-Mini Pro compressed source files.
Those marked "[F]" were not compressed via JPEG-Mini.
The numbers within the filenames indicate compression level in Publisher export dialog.
 
[F] 4/25/2020  19:36     387,526,265  Boise2018-NOCOMP.pdf
[C] 4/25/2020  15:35     388,483,998  Boise2018_JPEG-NOcompr.pdf
(unexpected result; export with uncompressed JPEGs is smaller, somehow)

[F] 4/25/2020  19:39     244,546,273  Boise2018-Compr100.pdf
[C] 4/25/2020  16:28     210,858,408  Boise2018_JPEGcompr-100.pdf

[F] 4/25/2020  19:43     210,600,433  Boise2018-Compr99.pdf
[C] 4/25/2020  16:23     182,725,987  Boise2018_JPEGcompr-99.pdf

[F] 4/25/2020  19:46     173,309,855  Boise2018-Compr98.pdf

With source files compressed with JPEG-Mini and a PDF compression setting of 90,
the PDF's size dropped to ~83 MB.

Switching for a couple of tests to PDF/X-1a:2003 produced crazy-large files and I gave up on that idea quick-like.

In another discussion about this, someone mentioned using Acrobat to optimize his PDFs' sizes. That's an expensive solution for someone like me who isn't in the trade. Perhaps there's other software specialized for that purpose and without all of Acrobat's bells and whistles. Then again those kinds of tools might produce optimized files that aren't suitable for press.

[lacerto]

(...)

[MikeA]

Thanks for the further information. If optimization could foul up the print job, I'll live without it. I just now uploaded the PDF to the on-demand book company and while it wasn't lightning-fast, it didn't give me fits. The export settings I used — PDF 1.4, downsampling on, JPEG compression=99 — produced a file just under the maximum allowed (300 MB). If I'd used the JPEG-mini files I could probably have added a few more images and used quality=100. But this is good enough for now. Well, assuming the book doesn't look like dog-meat-on-a-stick when they're done with it. : ) It's been a long haul learning Publisher from the ground up and acquainting myself with font-related tools I haven't had to think about for years. Thanks a lot for your advice along the way—it was extremely helpful.

[R C-R]

9 hours ago, MikeA said:

I take it this means that the max quality setting (100) still involves re-compression of the images. All along I'd thought 100 means: no compression. Apparently not.

Yes, even at the maximum "quality" setting the standard JPEG discrete cosine transform (DCT) encoder uses a form of lossy compression.

[MikeA]

On 4/26/2020 at 12:31 AM, Lagarto said:

Here is a comparison of quality difference (original left and optimized right, at 300% zoom level):

Interesting. Depending on my OCD level for the day, I'll think either that the differences are negligible or that they're very obvious. Then again 300% is a pretty extreme enlargement.

[lacerto]

(...)