Added JPEG quality option parameter (-c jpg_quality=n)

[tleegwater]

I needed to be able to specify the JPEG quality level in PDF output files, so I made this parameter optional. Default JPEG quality will still be 85.

[amitdo]

bce2cd5f331b66

[amitdo]

If my memory serves me well, @jbreiden (author of the pdf renderer code) didn't like this feature.

[tleegwater]

Ah, I see, d0cb107 says so.

http://web.archive.org/web/20160626112224/https://code.google.com/p/tesseract-ocr/issues/detail?id=1285

http://web.archive.org/web/20160626112223/https://code.google.com/p/tesseract-ocr/issues/detail?id=1300

My addition only handles JPEG quality where @jbreiden had an issue with all the extra encoding parameters.

[amitdo]

https://web.archive.org/web/20150413012101/https://code.google.com/p/tesseract-ocr/issues/detail?id=1300
@jbreiden commented:

I don't have a strong opinion about tessedit_pdf_jpg_quality. It might cause some confusion. But at least it will be harmless.

[jbreiden]

Let's talk about this for a bit. Tesseract's PDF module tries really hard to inline images instead of transcoding them, which means that the JPEG quality parameter should be rarely used. @tleegwater can you tell us what sort of image files you are feeding to Tesseract? Some sort of TIFF? Maybe attach one if possible?

I mainly want to make sure we don't have an accidental transcode situation.

[tleegwater]

I'm feeding TIFF file to Tesseract. And I'm aware that as soon as I feed it JPEG, or something else that's supported, Tesseract will try to inline the inputfile.
But even if someone would feed JPEG and set the quality parameter to 50, Tesseract still inlines the image without transcoding to q=50, or q=85 for that matter.

[jbreiden]

That's fine, and I'm fine with this changelist.

Can you please tell me what flavor of TIFF you are working with? Uncompressed? LZW? Pack? Deflate? JPEG? CCITT Group 4? You can find out with tiffinfo from libtiff, or identify -verbose from ImageMagick. I'm considering supporting a few more image format for inline, and this might influence prioritization.

[tleegwater]

We use no compression at all for our TIFF's We are generating them ourselves so there's no chance we'll get anything other than Uncompressed.

[jbreiden]

Got it. Please be aware that the built in JPEG encoder is standard libjpeg. If you ever want more precise control (for example, turning off chroma subsampling) a fancier encoder like Guetzli, then JPEG encode the images before feeding to Tesseract.

https://research.googleblog.com/2017/03/announcing-guetzli-new-open-source-jpeg.html

[amitdo]

I think the API break is unnecessary.

bool TessPDFRenderer::AddImageHandler(TessBaseAPI* api)
can use api->GetIntVariable("jpg_quality", &jpg_quality);

[zdenop]

Thanks for pointing to this issue. What it better for maintaining compatibility of C-API:

1. TESS_API TessResultRenderer* TESS_CALL TessPDFRendererCreate(const char* outputbase, const char* datadir, BOOL textonly, int jpg_quality=85);
2. or to define TessPDFRendererCreate (the old call) and TessPDFRendererCreate2 (new call with jpg_quality )?

[amitdo]

It's not just C API break, it's also a C++ break.
There are at least 2 bindings that use the C++ API directly (Python, R).

[amitdo]

As said. I think the C++ code can be changed to not break previous API.
Then you can revert the C API change.

[zdenop]

Ok. I got it. First I think about extending API, but it is not need because jpg quality is handled by tesseract parameter....

[zdenop]

Done. Please check.

[amitdo]

I didn't test it, but the change LGTM.