2006.12.05 20:14 "Re[2]: [Tiff] ITULAB JPEG support", by Jean-Yves Le Ridant

Joris a écrit:

More and by hell, we have a real language problem here. I really don't understand what you're saying.

I've been toying with the idea of adding a small interface on top of TIFFReadEncodedXxx. This small interface should be named TIFFReadSubsampledXxx. It will first call TIFFReadEncodedXxx, and if subsampling is in effect, it will de-subsample the data. It should be a valid call even if no subsampling is in effect, in which case it returns immediatelly after calling TIFFReadEncodedXxx.

This way, people will not have to worry about subsampling if they just remember to call TIFFReadSubsampledXxx where they used to call TIFFReadEncodedXxx. It is in the proper position, i.e. it is independent of the compression or color mode used, works on JPEG compressed YCbCr as well as on flate compressed CIELAB or whatever.

Is this what you are asking after?

Understand I dont (properly saying) "ask" for something else than a "responsable" for jpeg in tiff affairs.

:-)

What I post here must me more understood as remarks from what I experienced when trying to add some fonctionnality to "libjpegtiff", for my own use, and with in mind to be "contribuable" as possible, so "less change being beautifull".

( note that sources of what I place somewhere in the web during last thread are "a un ou deux bugs près" what I posted in bugzilla bazurk. :-) And, another point: I'm totally unaware of "how things goes" in evolutions of a packages like libtiff or jpeglib.....

My first impression is that this way may turn fairly complicated. Libjpeg ( except with some patch I did for it ) don't know anything about tiff's interleaved subsampled line, scanlinesize... ( note that such concept is "per se" questionnable ) So as soon as you deal with subsampled data, you have to use the "raw" interface of libjpeg, then buffer yourself an "imcurow" ( something as planar separate represention but with not exactly same alignement a tiff file would have ), and then, you'll had add a buffer level produce upsampled data as soon as you have a vsamp of 2....

To get YCbCr for client...

At the end, you dispatch/duplicate libjpeg, in both tif_jpeg.c and clients apps...

Ahma... :-))), and caricaturing, with goal of spreading files for as most reader/viewvers as possible, and simplifying tif_jpeg and surrounding, facing to disaccordance of IFD and what codec return: "correct" IFD when opening the file from what codec will return, and let wizzards use low-level functions to get subsampled data :-)))

--
Jean-Yves Le Ridant.