Motion JPEG 2000 (compression of grayscale images)
Mathieu Malaterre
I have been trying to search though multiple documents but I could
not find an answers to my question. Has there been any progress in
supporting Motion JPEG 2000 in DICOM. I went to the j2k-medical group
on yahoo (*), but there is only spam. The last mail is dated from:
Feb 2003:
http://groups.yahoo.com/group/j2k-medical/message/5
As a side note I am looking for a good compression algorithm for
grayscale images. JPEG 2000 lossless is giving me a 2:1 ration, which
is ok. But I would really benefit from the motion compression since my
dataset is basically a recording of a moving object. Has anyone played
a little with MPEG compression and quantified the data loss ?
Thanks
Mathieu
(*) This is the reference mentionned in:
http://www.jpeg.org/public/wg1n2883.pdf
David Clunie
DICOM already has an alternative to Motion JPEG 2000, and that
is the ordinary one JPEG 2000 "image" per-frame multi-frame image,
so we were not very excited by the existence of the MJ2K effort,
since it envisaged the prospect of having to support an additional
transfer syntax that duplicated the functionality that we already
had, or to transcode back and forth between one and the other.
There is no motion prediction in Motion JPEG 2000, only
successive completely independent J2K frames. The MPEG half
of ISO/IETC JTC1/SC29 group were very unhappy with the suggestion
of the JPEG half doing any work on motion, which they consider
to be their turf.
I am not sure if MJ2K has really been implemented or not; I got
the impression that the digital cinema folks were interested in
it.
MPEG has really only been used in medical imaging in the
context of cardiac ultrasound (echo), and in visible light
(endoscopy) applications, where the de facto standard for
quality previously was VHS tape (a pretty low standard).
Whether or not it was suitable for other types of imaging
or not would depend a lot on the modality and task.
David
PS. The j2k-medical group is not, as far as I am aware, terribly
active, nor terribly representative of medical imaging folks,
either users or vendors. DICOM does have an active formal liaison
with the main JPEG 2000 body, however. Of late, we have been more
interested in working on JPIP and 3D JPEG 2000.
Mathieu Malaterre
Thanks a lot for your detailed answer. I did not realized MJ2K was
indeed exactly encapsulating j2k frame one after the other. I was
confused by some post on kakadu newsgroup:
http://groups.yahoo.com/group/kakadu_jpeg2000/message/3889
kdu_v_compress is for Part 3 (motion JPEG2000)
Volume compression with exploitation of redundancy between
slices can be achieved using "kdu_compress". Take a look
at the advanced examples in "Usage_Examples.txt".
I will ask David Taubman about that.
Thanks again for your help !
Mathieu
Ps: I also discover at the same time an ongoing effort to implement
JPEG2000 - 3D by LPI, on top of OpenJPEG.
David Clunie
Compression of volumes is completely different.
Currently it uses the component transform defined in JPEG 2000
Part 2, and is quite effective for volumetric CT and MR, and
that has been added as a transfer syntax to DICOM.
There is also a JPEG 2000 3D effort, to allow rate allocation
in addition to the wavelet transformation.
David
Mathieu Malaterre
MJ2K does not do motion compensation. Anyway I am getting closer to my
goal, one bug remains:
http://groups.yahoo.com/group/kakadu_jpeg2000/message/4143
It looks like kakadu is having a hard time with simple 8bits data.
Mathieu