2001.10.05 21:21 "Re: OJPEG & Wang Images (long)", by Ross Finlayson
I have yet to see a standard that makes it unnecessary to support an old standard (where the old standard was actually in use)...
You are referring to tiff 6.0 jpeg compression as a 'standard'. This is exactly where you are wrong, in my opinion. It is not a standard because
- there are several possible interpretations that are mutually incompatible
- it is overwritten by technote #2
- it is in contradiction with the encapsulating standard (tiff standard says no offsets in data)
I'll add a quote, but since I'm not sure that he likes to be quoted (by an impolite rebel such as me), I will not mention his name. Residents in the list know who said it though, and respect the man (like I do).
'This is not a standard - it is a tower of Bable'
I think with the TIFF, instead of talking about annotations specifications that it is about mixed raster content, in terms of TIFF-FX, etcetera, what that is about is having separate regions of the image with different spectral characteristics, for example "full-page" TIFF/IT-8 FP image files containing the TIFF/IT files. In a way, those are already quite specified. I'm not quite sure what is the color registration method, except it just stores a reference to the ICC profile like the other software. Wang had an annotation extension that was using a private tag (?) and stored the data of the annotations, similarly to how various colorimetry fields and other extensions of the image format are already in daily use and acceptably, normatively specified.
There are many private uses of TIFF.
In a way, its about considering, with the file changes of an upgrade image format, exactly how to specify its compatibliity with TIFF. That is mostly about the file data organization of the image data, where TIFFs use 32-bit offsets, and as well the organization of the data in the TIFFs is sequential and location sensitive, in the file. When the file is to be accessed many times, then it is almost required for efficiency to use a database of the file offsets in the file, or at least a file containing that information for the software to use. To store that kind of data in the TIFF files, then what might happen would be that software that translates and doesn't support that feature would invalidate the offsets.
In a way, what you want for TIFF is a tiling infrastructure where the TIFFs are tiles. There are already specifications for tiles in TIFF, what this is about is having the reference space with the TIFFs which can have up to certain offsets being tiles. Then, it is about having a flexible datastore, for example how the tiles are stored in a database or in running or serialized program logic, or a file. That is mostly to allow that all the tools which use TIFFs would be able to simply use the TIFFs, or where the datastore of all the information has TIFF format images of the data.
For example, there is GeoTIFF, with Mercator referencing, where the software logic can work with the extension image data. That is. JPEG2000 uses a component registration system in that manner, in a way, to consider using JPEG2000 besides wavelets for compression of TIFF images basically considers that TIFF is a specified, established standard, in use for its task.
I think it is good to add support for the varieties of TIFF to libtiff, the free TIFF reference software.
About JPEG2000, see my web page with information about JPEG2000, http://www.apexinternetsoftware.com/jsdoc/html/.
I posted an empty implementation for the OJPEG to this list, libtiff, tif_ojpeg.c, it's in the archive, just so that libtiff reads the OJPEG field types. Then, I use TIFF Read and Seek on the TIFF files, where the OJPEG fields are mostly containing offsets to the JPEG data. Then, there is a discrepancy between TIFF OJPEG and JPEG interchange format, for much of the interpreter software which uses the JPEG restart markers, RSTm, they are optional in the OJPEG. Then, I have seen other "Old JPEG" files which do not adhere to the specification.
Mostly, the files adhere to the specification. There are many selective uses. Almost completely, the files that use standard fields are standard in their use of those fields, except perhaps LEADtools and Accusoft ImageGear, HP, that I have seen, as a developer working on software for "TIFF files" and only in terms of their "OJPEG" files, except for the LEADtools "TIFF" files labelled JPEG. That is about how the Wang or PixTran, etc., JPEGs appear to be actually adhering to the specification. Almost a hundred percent of the image files conform directly to the specification.
company Apex Internet Software