TIFF compression options

[eL_PuSHeR]

Do ScanTailor just supports LZW compression in TIFF? What about others (e.g. ZIP)?

Not a big deal and working with LZW equals less compatibility issues. But I am just curious...

[Tulon]

I always try to minimize the number of preferences, therefore I choose to have just one output format. When deciding on the format, the following issues were taken into account:
* Can't be lossy, so no JPEG.
* LizardTech's DjVu encoders didn't support PNG or TIFF Deflate (aka ZIP), so those are out too.
* TIFF G4 would be good for B/W output, but Photoshop gets its aspect ratio wrong.

So, TIFF LZW was the only option that doesn't have compatibility issues. Fortunately, by the time I started developing Scan Tailor, the patent on LZW compression had already expired.

[eL_PuSHeR]

It seems reasonable. Thank you very much for your reply.

[knappen]

A very, very elementary question:

Is TIFF more effectively compressed in a PDF conversion than JPG? I have tended to go through all the color and greyscale images and saved them as JPG as I thought this would result in a smaller end product before making the PDF with Adobe Pro, but suddenly I'm not sure that this is the way to proceed after all ...

[Misty]

No, compressing the TIFFs to another format before using Acrobat Pro won't improve your compression. PDF uses several different compression methods, including JPEG and the superior JPEG2000, so it's actually better if you haven't used a lossy compression format before sending it to Acrobat. JPEG is also a poor format for Scan Tailor text compression, because it's pure black and white. Acrobat will select a better compression automatically.

If your TIFFs have both illustrations and text on the same page, you'll get the best compression by using a script that can separate layers like Strider1551's djvubind for DjVu, or my own PDF maker script that I'll be releasing in the near future.

[spamsickle]

Is your PDF maker going to generate separate text and image layers? I assume you won't be doing real text with a font (so I still don't need to bother with an OCR step), but I also assume a text layer could use run length encoding or some other efficient compression on the binary portion of the image.

[Misty]

That's right. It separates each page into two layers, one binarized text and one colour/greyscale image. The text is the scan as produced by Scan Tailor, not real text with a font. The text is compressed using lossy JBIG2, which is really efficient for pure black/white images like text. I'm using JPEG2000 for images. Text is used at full-res, but I'm offering the option of selecting a separate DPI for the text so you can get smaller PDFs by using screen-size downsampled images - 100DPI by default.

The first version won't include OCR; that will need to be done using Acrobat as a separate step by the user. I'm looking at the possibility of adding Tesseract to do OCR in a later version.

[knappen]

Thanks, misty. I guess I have wasted time sorting out and converting illustration images that would have been better left alone.

Looking forward to the script.

Does anyone have a hot tip on a guide to getting a good view of different image formats and their respective use for cases like this?

[Misty]

A guide like that would be handy! I'm not sure if there is one. I'm planning to do a little glossary on the different file/compression formats for my blog after I finish the post I'm putting together on camera selection.