Compression - Reverse Engineering: Variable-length encoding of differences of

Author

Reverse Engineering: Variable-length encoding of differences of

mathieu

2007-11-15, 7:04 pm

Hi,

[crosspost from ffmpeg-users as suggested]

I am currently involved in an open source project: GDCM. This is
basically a toolkit to read/write DICOM medical images. We have in the
past been able to work around minor bugs from private vendors (wrong
huffman table, jpeg predictor stored on 16bits instead of 32...). But
lately we have been facing a greater challenge: complete unknown
compression scheme. For reference an example DICOM file can be found
here:

http://gdcm.sourceforge.net/LosslessRice/I160

and the extracted raw compressed stream:

http://gdcm.sourceforge.net/Lossles...60_07a1100a.raw

This represents a 512x512 16bits grayscale image.

According to a comment in the DICOM file I can see 'LOSSLESS RICE'
and from (reliable?) sources on the web it could be related to US
Patent 4809350...

So where do I start if I were to decompress this file ?

Thanks for your time,
-Mathieu

Jim Leonard

2007-11-15, 7:04 pm

On Nov 15, 10:03 am, mathieu <mathieu.malate...@gmail.com> wrote:
>
> According to a comment in the DICOM file I can see 'LOSSLESS RICE'

Golomb-Rice coding perhaps?

Sponsored Links