file size increase for pdf/a

OCRmyPDF is really marvelous! Thanks!

I have one question regarding output file size: Unless explicitly selecting pdf as output type, I have quite large file sizes (~4x) after “ocrmypdf in.pdf out.pdf”. The pages are scanned text, i.e. actually there are no gray pixels only black or white ones. Only “–output-type pdf” keeps the file size similar.

For the first page (the others are similar) “pdfimages -list in.pdf” gives:

page   num  type   width height color comp bpc  enc interp  object ID x-ppi y-ppi size ratio
--------------------------------------------------------------------------------------------
   1     0 image    2697  4533  icc     1   1  ccitt  no        17  0   600   601 88.3K 5.9%

out.pdf results in:

page   num  type   width height color comp bpc  enc interp  object ID x-ppi y-ppi size ratio
--------------------------------------------------------------------------------------------
   1     0 image    2697  4533  rgb     3   8  image  no        12  0   600   601  385K 1.1%

Even --optimize 3 results in double file size for out.pdf (saved as pdf/a):

page   num  type   width height color comp bpc  enc interp  object ID x-ppi y-ppi size ratio
--------------------------------------------------------------------------------------------
   1     0 image    2697  4533  gray    1   1  image  no        34  0   600   601  203K  14%

Is a conversion obligatory for pdf/a? Or is there a way to keep the original image type AND generate pdf/a?

[femifrak]

Thank you for the explanation. But one thing I don’t understand:

with -output-type pdf (unless the “Image processing” settings are turned on), the raster image will never make it into the final PDF.

Did you mean “without” instead of “with”? With “-output-type pdf” I get small outfile sizes. Therefore, I think this option makes the original PDF images the final PDF.

I would think that most of the scanned PDFs contain old documents, many of which are only b/w. That is, I don’t think b/w is such a rare special case, don’t you? What do you think of an option to convert gray images to b/w for output? Gray images would be better for tesseract (?) and b/w output would be better for reading and for file size. This would be the same reasoning as for --clean-final, but --clean-final doesn’t convert to monochrome. (Apart from the programming effort, of course …).

change compression after quantizing. That is something I can add eventually.

very good idea 😃

[aspik]

I know this might be not the right place, but I didn’t want to create a new “issue”.

I don’t intend to offer save as b/w because some workflows (mine for example) is mostly b/w with a few color images per file, and I’d rather not introduce an option that effectively ruins the output if you use it on the wrong file.

Could you please tell a bit more about your workflow? I have documents which contains just black text with a small color graphic (corporate logo). Additionally one site has signatures which are made with a blue pencil. What would be the best way to scan and process this kind of documents (contracts)? I don’t want to lose the color information.

Currently I’m scanning it as color text with 600dpi (using VueScan). After passing it through ocrmypdf the size did not reduced much (~700KB). When using the jbig2-lossy compression 3 the size was halved (14MB -> 7MB for an 8 pages document). I’m perfectly fine with the size in case of storing it locally.

However, sometimes I want to send this kind of document per e-mail and in this case even the 7MB is not optimal. I would like to convert it to b/w as it does not matter if the corporate logo is red-blue or just black. How would that be harmful if there would be a save/convert as b/w option?

Maybe I’m missing something here and there is a different way to reduce the file size?

Thanks for the excellent app!