[Openexr-user] Details in PIZ compression

[Jack Mak]

Hi Experts,

How does the PIZ compression work? Where can I find reference that
discusses it in detail? (I checked out the documentations on the
OpenEXR website, but they does not help.)

Many thanks

Jack

[Florian Kainz]

For the details of PIZ compression, you'll have to
look at the source code, but here's a brief overview:

The image is split into blocks that are 32 pixels high
and as wide as the image. The blocks are split into
individual channels. The bit-patterns that comprise
the half values in each block are re-interpreted as
16-bit integers, and the integers are run through a
lookup table that shapes their distribution such that
comb-like histograms, with spikes separated by wide
gaps, are avoided. A Haar wavelet transform is applied
to the output of the lookup table, and the output of
the wavelet transform is compacted using a 16-bit
Huffman coder.

PIZ is mathematically lossless because each individual
step is reversible. The wavelet transform is reversible
because it operates on integers and has no roundoff error.

Florian

Post by Jack Mak
Hi Experts,
How does the PIZ compression work? Where can I find reference that
discusses it in detail? (I checked out the documentations on the
OpenEXR website, but they does not help.)
Many thanks
Jack
_______________________________________________
Openexr-user mailing list
http://lists.nongnu.org/mailman/listinfo/openexr-user

[Jack Mak]

Hi Florian,

Thanks for your explanation. I got the big picture. Before I look through
the source code, just wonder, are there any documents (papers, standards...)
give a detailed description of PIZ compression?

Best regards,

Jack

Post by Florian Kainz
For the details of PIZ compression, you'll have to
The image is split into blocks that are 32 pixels high
and as wide as the image. The blocks are split into
individual channels. The bit-patterns that comprise
the half values in each block are re-interpreted as
16-bit integers, and the integers are run through a
lookup table that shapes their distribution such that
comb-like histograms, with spikes separated by wide
gaps, are avoided. A Haar wavelet transform is applied
to the output of the lookup table, and the output of
the wavelet transform is compacted using a 16-bit
Huffman coder.
PIZ is mathematically lossless because each individual
step is reversible. The wavelet transform is reversible
because it operates on integers and has no roundoff error.
Florian

Post by Jack Mak
Hi Experts,
How does the PIZ compression work? Where can I find reference that
discusses it in detail? (I checked out the documentations on the
OpenEXR website, but they does not help.)
Many thanks
Jack
_______________________________________________
Openexr-user mailing list
http://lists.nongnu.org/mailman/listinfo/openexr-user

...

[Florian Kainz]

Unfortunately there isn't a document that describes PIZ (or any of the
other compression methods) in detail. If anyone wants to write such a
description, I'd be happy to put it on the web site, along with the
other documentation.

Post by Jack Mak
Hi Florian,
Thanks for your explanation. I got the big picture. Before I look
through the source code, just wonder, are there any documents (papers,
standards...) give a detailed description of PIZ compression?
Best regards,
Jack
For the details of PIZ compression, you'll have to
The image is split into blocks that are 32 pixels high
and as wide as the image. The blocks are split into
individual channels. The bit-patterns that comprise
the half values in each block are re-interpreted as
16-bit integers, and the integers are run through a
lookup table that shapes their distribution such that
comb-like histograms, with spikes separated by wide
gaps, are avoided. A Haar wavelet transform is applied
to the output of the lookup table, and the output of
the wavelet transform is compacted using a 16-bit
Huffman coder.
PIZ is mathematically lossless because each individual
step is reversible. The wavelet transform is reversible
because it operates on integers and has no roundoff error.
Florian

Post by Jack Mak
Hi Experts,
How does the PIZ compression work? Where can I find reference that
discusses it in detail? (I checked out the documentations on the
OpenEXR website, but they does not help.)
Many thanks
Jack
_______________________________________________
Openexr-user mailing list
http://lists.nongnu.org/mailman/listinfo/openexr-user

...

[Klaus Steden]

If there were a rudimentary set of notes to start with ... I remember
hearing it described in a seminar at SIGGRAPH a few years back, so I seem
to recall it involves matrix transforms and so forth, but the rest is a
bit fuzzy by now.

Post by Florian Kainz
Unfortunately there isn't a document that describes PIZ (or any of the
other compression methods) in detail. If anyone wants to write such a
description, I'd be happy to put it on the web site, along with the
other documentation.

Post by Jack Mak
Hi Florian,
Thanks for your explanation. I got the big picture. Before I look through
the source code, just wonder, are there any documents (papers,
standards...) give a detailed description of PIZ compression?
Best regards,
Jack
For the details of PIZ compression, you'll have to
The image is split into blocks that are 32 pixels high
and as wide as the image. The blocks are split into
individual channels. The bit-patterns that comprise
the half values in each block are re-interpreted as
16-bit integers, and the integers are run through a
lookup table that shapes their distribution such that
comb-like histograms, with spikes separated by wide
gaps, are avoided. A Haar wavelet transform is applied
to the output of the lookup table, and the output of
the wavelet transform is compacted using a 16-bit
Huffman coder.
PIZ is mathematically lossless because each individual
step is reversible. The wavelet transform is reversible
because it operates on integers and has no roundoff error.
Florian

Post by Jack Mak
Hi Experts,
How does the PIZ compression work? Where can I find reference that
discusses it in detail? (I checked out the documentations on the
OpenEXR website, but they does not help.)
Many thanks
Jack
_______________________________________________
Openexr-user mailing list
http://lists.nongnu.org/mailman/listinfo/openexr-user

_______________________________________________
Openexr-user mailing list
http://lists.nongnu.org/mailman/listinfo/openexr-user

...