Question about Lossless Transfer Syntaxes and Color Transformations.
Don Mimlitch
in RLE Lossless with a Photometric Interpretation (0028,0004) of
"YBR_FULL" and a Planar Configuration (0028,0006) of "1" (Planar).
If an Image Archive SSCP receives and Stores such images as RLE
Lossless, the standard seems to imply that if it forwards them
uncompressed, they must remain as "YBR_FULL". There are a number of
viewers still in use which do not properly display "YBR_FULL" images.
My question is whether the Image can be transformed to RGB (which is
marginally lossy since Floating Point Arithmetic is involved) without
assigning it a new SOP Instance UID and marking it as Lossy
Compression (0028,2110)? The Lossy Image Compression Ratio (0028,2112)
would be 1.00 in this case.
The transformation from Planar Configuration of One (Planar) to Planar
Configuration of Zero (Interleaved) is lossless and should be allowed.
David Clunie
Though the rule requiring a new SOP Instance UID is based
on changing the "interpretation", I would argue strongly
that whenever there is a non-lossless transformation of
the pixel data one should err on the side of changing the
UID, as in this case, even though the loss is small and
due to calculation error.
david
PS. I can't imagine anyone would ever use RLE and YBR for
XA images - surely only for color US.
donmi...@insiteone.com (Don Mimlitch) wrote in message news:<1214259b.03032...@posting.google.com>...
Don Mimlitch
Hi Dave
In a response to another question you gave the following quote from the Dicom
Standard:
The rule for when to change the SOP Instance UID for an
image is:
"If the pixel data of the derived Image is different from
the pixel data of the source images and this difference is
expected to affect professional interpretation of the image,
the Derived Image shall have a UID different than all the
source images." PS 3.3 C.7.6.1.1.2
In order to see how the conversion from RGB->YBR and YBR->RGB affected the
color transformation, I wrote a small program to convert:
R0 G0 B0 -> Y Cb Cr -> R G B
for all 16 Million combinations of RGB and then scored the differences:
score = (R - R0)^2 + (G - G0)^2 + (B - B0)^2
The Worst score for any RGB value was 3 (i.e. a case where each of the values
of R, G and B were off by 1 out of 256 possible values.
If I iterated the conversion one more time to see if the error accumulated:
R0 G0 B0 -> YBR -> RGB -> YBR -> RGB
The worst score for any RGB value was 8 (i.e. two of the components of RGB
were off by 2).
If I repeated this iteration 100 times the worst error score was still 8.
The point of all this is that I'm not sure that even the best Professional
Interpreter could visually distinguish such a small color variation.
Also, in my very limited understanding of Ultra Sounds, it seems that the
colors are computed by the modality to highlight flows and the choice of
colors depends on the manufacturer. So it would seem improper for any
interpreter to give meaning to fine changes in Hue even if they were visible.
This would seem to imply that the Profession interpretation would not be
affected by this transformation and thus a new SOP Instance UID is not needed.
Again my knowledge of Ultra Sounds is limited to the Mathematics and not real
world interpretation, so I would appreciate your comments on my assumptions.
Don