2011.03.29 20:17 "[Tiff] LibTIFF compression enhancements", by Stephan Busch

2011.03.29 20:17 "[Tiff] LibTIFF compression enhancements", by Stephan Busch

Dear Thomas,

it has been a while since we last talked.
I suggested a new lossless code for usage within your TIFF library.
Neither PackPNM nor the popular JPEG2000 has been implemented by now.

Today I also want to suggest a new codec.

It is fast. It is free. It is open-sourced.

BCIF by Stefano Brocchi

http://www.dsi.unifi.it/users/brocchi/

The caveat is that it supports 24-bit color images only by now.

The author would be interested in adding BCIF to your TIFF library.
What do you think?

Looking forward to your answer

Regards,
Stephan

________________________________________________________________________
> Hi,

>> So far I haven't heard of any TIFF libraries that use jpeg2000 algorithm= s,

>> but its sad that all those maintainers could not agree to use a unified= =20

>> version all libraries could read.

> I'm not quite clear what you mean by that. JPEG2000 is standardized, and > if TIFF tags for J2K are specified, you should be able to include it.=20

> However, I probably don't see the point. J2K defines its own file=20

> format, avoiding typical TIFF problems like the 4GB barrier, thus I=20 > wonder why anyone would want to do that.

>> LOCO-I is the algorithm behind JPEG-LS. It was designed a decade ago by= =20

>> HP Labs,
>> gives better compression than jpeg2000 in most cases and there are many=
=20
>> implementations;
>> some of them are open-sourced such as:
>> www.stat.columbia.edu/~jakulin/jpeg-ls

> While I certainly agree for lossless compression, just a note: Please do > not use the above implementation, it has a bug in bit depths > 8 and=20

> creates files that do not conform to the specs - it's an endian issue=20 > that can be easily fixed, but that is nevertheless there.

> JPEG-LS is certainly a nice tiny codec that shouldn't take up too much > time to implement (hopefully then correctly) but are there TIFF tags for = it?

>> In the case neither jpeg2000 nor jpeg-ls are implementable, I wanted to= =20

>> suggest
>> another codec that will be free and open-sourced soon. The algorithm is=
=20
>> called PackPNM at present
>> and it performs like jpeg-ls.

> I have never seen or heard about this one. I would prefer algorithms=20 > that are openly available, actually.

>> Also, in my opinion only the codec that you will chose might survive.=20 >> All high performance image codecs

>> have been designed but were never implemented for practical use.=20

> What do you mean by that? JPEG-LS and JPEG 2000 have been implemented=20 > for practical use for sure.

> So long,
> Thomas

Stephan Busch