Motion JPEG 2000 (compression of grayscale images) - DICOM

This is a discussion on Motion JPEG 2000 (compression of grayscale images) - DICOM ; Hello, 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 ...

+ Reply to Thread
Results 1 to 5 of 5

Thread: Motion JPEG 2000 (compression of grayscale images)

  1. Motion JPEG 2000 (compression of grayscale images)

    Hello,

    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


  2. Re: Motion JPEG 2000 (compression of grayscale images)

    Hi Mathieu

    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 wrote:
    > Hello,
    >
    > 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
    >


  3. Re: Motion JPEG 2000 (compression of grayscale images)

    David,

    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...0/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 wrote:
    > Hi Mathieu
    >
    > 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 wrote:
    > > Hello,
    > >
    > > 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
    > >



  4. Re: Motion JPEG 2000 (compression of grayscale images)

    Hi Mathieu

    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 wrote:
    > David,
    >
    > 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...0/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 wrote:
    >> Hi Mathieu
    >>
    >> 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 wrote:
    >>> Hello,
    >>>
    >>> 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
    >>>

    >


  5. Re: Motion JPEG 2000 (compression of grayscale images)

    Ok I see. Now it all make sense, except that I still cannot believe
    MJ2K does not do motion compensation. Anyway I am getting closer to my
    goal, one bug remains:

    http://groups.yahoo.com/group/kakadu...0/message/4143

    It looks like kakadu is having a hard time with simple 8bits data.

    Mathieu

    David Clunie wrote:
    > Hi Mathieu
    >
    > 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 wrote:
    > > David,
    > >
    > > 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...0/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 wrote:
    > >> Hi Mathieu
    > >>
    > >> 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 wrote:
    > >>> Hello,
    > >>>
    > >>> 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
    > >>>

    > >



+ Reply to Thread