YBR_FULL / YBR_FULL_422 / YBR_PARTIAL_422
Schlomo Aschkenasy MD
I'm looking for an explanation of the YBR_* image formats. I didn't find any
information on the Internet or in the DICOM manuals. Can someone tell me
what this format is all about or where I can find information about writing
a decoder?
Thanks a lot,
-Schlomo Aschkenasy
Zurich - Switzerland
David Clunie
Historically these were added when the Ultrasound supplement in order
to support lossy JPEG compression of color images (which in retrospect
wouldn't have been possible prior to that supplement since even though
the transfer syntaxes for JPEG lossy were present, there was no way to
signal the color space transformation in use ... it isn't in the JPEG
bitstream itself).
In practice, YBR_FULL_422 is what is (almost ?) always used, since it
corresponds to what people in the outside world do when they JPEG
compress color images (e.g. on the web). Specifically, the RGB color
values are transformed into YCbCr as per the equations (and this is a
lossy step), and the two chrominance channels are downsampled by 2 in
the row direction. This, for example, is what the IJG cjpeg utility
would do for you if not otherwise specified (you can mess with the
sampling with command line options). It corresponds to what is defined
for the "JFIF" file format to encapsulate JPEG bitstreams, which
later evolved into SPIFF (ISO 10918-3). See:
http://www.ijg.org/files/jfif.txt.gz
http://www.jpeg.org/public/spiff.pdf
I am not quite sure why the YBR_PARTIAL_422 is there ... I believe it
is intended for when the transformed captured video signal does not
occupy the full width of the 8 bits available, but I haven't got a
copy of ITU-R BT.601-1 to check this. To put this another way, I
gather that in 601-1, there were different transformations from RGB
data as opposed to sampled video signals. I don't know if anyone has ever
used YBR_PARTIAL_422 for JPEG in DICOM objects. Sorry I can't be more
specific, but I am appallingly ignorant when it comes to the subject
of digitized video.
YBR_FULL is defined in Part 11 for use with RLE compression on ultrasound
media, with no downsampling in the chrominance channel - it still isn't
lossless though, due to the floating point calculations, so I don't know
if it is actually used in practice.
FYI, in Sup 61 for JPEG 2000, there are some more color space transformations
defined to correspond to what is recommended in the JPEG 2000 standard, one
reversible, and one not.
david
--
David A. Clunie mailto:dcl...@radpharm.com
Director, Technical Operations http://www.radpharm.com/
RadPharm Work 609-936-2600 Fax 609-936-2601
103 Carnegie Center Drive #300A Home 570-897-7123 Fax 425-930-0171
Princeton NJ 08540 http://www.dclunie.com/