Successive compressions and Source Image Sequence - DICOM

This is a discussion on Successive compressions and Source Image Sequence - DICOM ; Both Lossy Image Compression Ratio (0028,2112) and Lossy Image Compression Method (0028,2114) are explicitly are multivalued to be able to convey a((n) hopefully, short) history of successive lossy compressions. I *think* that multiple items in Derivation Code Sequence (0008,2112) can ...

+ Reply to Thread
Results 1 to 2 of 2

Thread: Successive compressions and Source Image Sequence

  1. Successive compressions and Source Image Sequence

    Both Lossy Image Compression Ratio (0028,2112) and Lossy Image
    Compression Method (0028,2114) are explicitly are multivalued to be
    able to convey a((n) hopefully, short) history of successive lossy
    compressions. I *think* that multiple items in Derivation Code
    Sequence (0008,2112) can also convey this history.

    I'd like to check my assumptions about For Source Image Sequence
    (0008,2112). If an image has had a number of SOP Instance UIDs in its
    life, it would be very helpful to know that history, for example, to
    resolve a reference to an earilier one, or when compiling an Alternate
    Representation Sequence (0008,3001). I wasn't certain whether the
    wording "the source images used to create the Derived image" supports
    that interpretation, or if this would be a misuse of the sequence.
    Part of my uncertainty is that Source Image Purposes of Reference
    (Context ID 7202) includes "Uncompressed predecessor" (DCM,121320), but
    nothing along the lines of "Previous compressed version".

    Q: Is it appropriate to accumulate all of the prior SOP Instance UIDs
    in Source Image Sequence if a lossy compressed image is subsequently
    re-lossy-compressed?

    Q: If so, can I use "Uncompressed predecessor" (DCM,121320) for the
    Purpose of Reference Code?

    Thanks VERY much,
    Dan Konigsbach


  2. Re: Successive compressions and Source Image Sequence

    Hi Dan

    dkonigsbach@dominator.com wrote:

    > Both Lossy Image Compression Ratio (0028,2112) and Lossy Image
    > Compression Method (0028,2114) are explicitly are multivalued to be
    > able to convey a((n) hopefully, short) history of successive lossy
    > compressions. I *think* that multiple items in Derivation Code
    > Sequence (0008,2112) can also convey this history.


    The number of items allowable for Derivation Code Sequence (0008,2112)
    is not specified, though it should be. This definitely needs a CP.

    If multiple items are present, then their meaning in relation to one
    another is not described, and therefore does not imply successive
    application of derivation steps (or anything else). A CP should
    define this one way or the other or emphasize that it is not defined.

    > I'd like to check my assumptions about For Source Image Sequence
    > (0008,2112). If an image has had a number of SOP Instance UIDs in its
    > life, it would be very helpful to know that history, for example, to
    > resolve a reference to an earilier one, or when compiling an Alternate
    > Representation Sequence (0008,3001). I wasn't certain whether the
    > wording "the source images used to create the Derived image" supports
    > that interpretation, or if this would be a misuse of the sequence.
    > Part of my uncertainty is that Source Image Purposes of Reference
    > (Context ID 7202) includes "Uncompressed predecessor" (DCM,121320), but
    > nothing along the lines of "Previous compressed version".


    There is no specified succession semantics for multiple items in the
    Source Image Sequence (0008,2112), and the presence of multiple items
    usually means that all those images were combined to make this image
    (e.g. multiple source images to make an MPR or MIP).

    > Q: Is it appropriate to accumulate all of the prior SOP Instance UIDs
    > in Source Image Sequence if a lossy compressed image is subsequently
    > re-lossy-compressed?


    I don't think many receivers would be able to recognize this pattern of
    usage, assuming there are any receivers that care, though it is not
    explicitly forbidden.

    The use of the Purpose of Reference Code Sequence (0040,A170), which is
    a relatively recent addition to the standard, within the references could
    make the intent more clear. But there is still no way of indicating the
    order of succession and distinguishing when multiple images are combined
    to create this image as opposed to multiple images being derived from one
    another in a particular order.

    > Q: If so, can I use "Uncompressed predecessor" (DCM,121320) for the
    > Purpose of Reference Code?


    There is currently no code specifically for recompression of a previously
    lossy compressed predecessor. One could be added if you think it is
    important.

    In general, two aspects of maintaining "history" information within the
    image objects themselves are relatively weak; one as you observe is the
    history of the pixel data; the other is the history of the entire object,
    such as when other non-pixel data attributes are modified, removed, added
    or coerced.

    David


+ Reply to Thread