Am 23.08.2017 um 18:40 schrieb Andreas Lehmkuehler:
readded dev@pdfbox
Am 22.08.2017 um 19:14 schrieb Jörg Henne:
Am 19.08.2017 um 17:07 schrieb Andreas Lehmkuehler:
The following files don't have a license header:
Good catch. Tracked as https://github.com/levigo/jbig2-imageio/issues/46
What about the binary test files in src/test/resources/? I assume
their license is cleared as well, isn't it?
That's what I assumed as well, but upon re-checking, things no longer
seem to be so clear. I'm tracking this question as
https://github.com/levigo/jbig2-imageio/issues/48
Maybe you guys can help me with this problem or let me know how you
deal with it.
Is there any jbig2-viewer available?
In theory, yes, for example XnView supports JBIG2 via jbig2dec.exe. In
reality, support for the various cases covered in the test suite is
rather spotty: many of the images cannot be decoded with XnView. So,
strange as it might seem, I don't know of any reliable stand-alone JBIG2
viewer.
However, obviously those images can be decoded using the plugin. I've
attached PNG versions of them to a comment on the above issue:
https://github.com/levigo/jbig2-imageio/issues/48#issuecomment-324556311
Are these testfiles somehow special, do they trigger some special
processing within the plugin or are they just a bunch of jbig2 files
and could be replaced by others
JBIG2 isn't quite as simple as, say, PNG. There are several entropy
coding options (Arithmetic/MQ, Huffman) several different segment types
and several ways to maintain, refine and reference shape dictionaries.
Therefore there a large number of code paths need to be covered in the
tests. Since it is rather hard to generate all those possible
combinations (no single encoder library will use all of them) the
refrerence library provides (provided?) a convenient way of achieving
decent test coverage.
The files seem to fall into three categories:
1. Files from the original test suite. While the copyright status of
the file isn't problematic, the status of the content seems to be
muddy in some cases.
- Files containing representations of public U.S. government
documents should be in the public domain:
https://en.wikipedia.org/wiki/Copyright_status_of_work_by_the_U.S._government
- The same applies to representations of U.S. patents:
https://en.wikipedia.org/wiki/Copyright_on_the_content_of_patents_and_in_the_context_of_patent_prosecution
- 004.jb2 and 005.jb2 seem problematic but may be covered by some
exemption.
- amb.bmp no idea
amb.bmp seems problematic as it looks like a promo photo of Ally
McBeal aka Callista Flockhart.
You seem to be more up to speed regarding TV characters. I certainly
didn't recognise the person in the photo :-)
Losing this image would be bad, though, since it is the only halftone
region sample bitstream in there.
2. Files provided to us with the permission to use them for testing
purposes
201231100*.jb2 is the only case, seems to be a public U.S.
document anyway and therefore in the public domain. I have not
contacted the original provider of the files for the simple reason
that his or her e-mail address has been lost when the Googlecode site
went into archived state. >
3. Files with content so trivial that copyright should not be an
issue, i.e. fragments of bitstreams, isolated segments, trivial test
images
This isn't a question of copyright but of license and/or privacy.
The files in this category are sampledata_page(1,2,3).jb2. The content
is obviously not a matter of privacy. Regarding the license I am
currently asking around whether anyone still knows where this came from
(unfortunately we lost some very early RCS history from before we
open-sourced the component).
Jörg
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
