PDF XChange Forum

Hello All,

I am a legal user of PDF-Tools 4 and I evaluate the PDF-Tools 4 SDK.
I wish to know if the new PDF-Tools 5 could better compress pages with just one B&W bitmap using JBig2 options.

It seems that, using any PDF-Tools (version 4) JBig2 option, a B&W bitmap is not very well compressed, comparing with other tools as the Jbig2enc freeware (https://github.com/agl/jbig2enc).
Here enclosed a zip archive with 5 files :
- a 1 Tiff page from US from US Patent Office (image of 2320x3408 px): 05307062-001.tif (46,141 bytes)
- a 1 Pdf page converted using jbig2enc.exe: 05307062 -001.pdf (27,938 bytes)
- a 1 Pdf page converted using PDF-Tool 4 (opt. symbols): 05307062-001jb2sym.pdf (40,710 bytes)
- a 1 Pdf page converted using PDF-Tool 4 (opt. standard): 05307062-001jb2std.pdf (36,390 bytes)
- a 1 Pdf page converted using PDF-Tool 4 (opt. limit): 05307062-001jb2lim.pdf (36,390 bytes)

Could we hope a better compression rate using PDF-Tools 5 for this kind of conversion?

Best regards
Gérard

Hello Gerard,

I just spoke with the Viewer project leader and he told me that we will try to make improvements in the compression algorithms for JBig2, and he was also surprised that symbol compression is not producing the smallest files.

He said that he can't promise any time frame when this could be done, but we've created a ticket in our internal system:
#1488: Better compression rate of jbig2 format for PDF-Tools SDK 5?
So that we can track this case and make the necessary improvements.

Best,
Stefan

Hi Stefan

Many thanks for your reply.

"...he was also surprised that symbol compression is not producing the smallest files..." me too, because "Symbols" is the best algorithm to compress a composite B&W bitmap (it proceeds an "OCR-like scan" of all the patterns found in a page and builds a common dictionary of these symbols for all the pages of a PDF file).

But I wonder if you use the very good but slow "arithmetic coding" algorithm (> 30 % better) or the quicker but not very good "generic coding"?

Gérard

Hi Gérard,

I am not sure of the exact algorithms used, but while discussing it our project leader told me that he is considering some further analysis and better pattern selection or any other possible improvements. As for the actual speed vs compression algorithm used - I believe that the three different options should be covering the spectrum but will need to speak with him again to confirm that if needed.

Best,
Stefan

Hi Gérard,

I have some update in the ticket - and it's that while our devs will be implementing improvements in this respect, they won't happen in the very next few builds and this will be a bit of a longer term item in our "TO DO" list.

Best,
Stefan

Hi Stefan

Many thanks for this news. We wait...

Best regards

Gérard