R/nrrd-io.R
write.nrrd.Rd
write.nrrd
writes an array, vector or im3d object to a
NRRD file. When x
is an im3d
object, appropriate spatial
calibration fields are added to the header.
write.nrrd.header
writes a nrrd header file.
write.nrrd.header.for.file
makes a detached NRRD
(nhdr) file that points at another image file on disk, making it
NRRD compatible. This can be a convenient way to make NRRD inputs for other
tools e.g. CMTK and also allows the same data block to pointed to by
different nhdr files with different spatial calibration.
Data to write as an array
, vector
or
im3d
object.
Character string naming an output file (a detached nrrd header
when file
has extension 'nhdr').
One of three supported nrrd encodings ("gzip", "raw", "text")
The data type to write. One of "float","byte", "short", "ushort", "int", "double"
List containing fields of nrrd header - see Header section.
One of "big" or "little". Defaults to .Platform$endian
.
Optional name of separate file into which data should be written (see details).
Path to input and output file for
write.nrrd.header.for.file
. If outfile
is NULL
(the
default) then it will be set to <infilestem.nhdr>
.
NRRD files can be written in detached format (see http://teem.sourceforge.net/nrrd/format.html#detached) in which a text nhdr file is used to described the contents of a separate (usually binary) data file. This means that the nhdr file can be inspected and edited with a text editor, while the datablock can be in a completely raw format that can be opened even by programs that do not understand the NRRD format. Furthermore detached NRRD header files can be written to accompany non-NRRD image data so that it can be opened by nrrd readers.
If file
has extension .nhdr
or datafile
is
non-NULL, then write.nrrd
will write a separate datafile. If
datafile
is set, then it is interpreted as specifying a path relative
to the nhdr file. If datafile
is not specified then default
filenames will be chosen according to the encoding following the
conventions of the teem library:
raw '<nhdrstem>.raw'
gzip '<nhdrstem>.raw.gz'
text '<nhdrstem>.ascii'
When a detached NRRD is written, the
datafile
can be specified either as relative or
absolute path. Relative paths are strongly recommended - the best
place is right next to the datafile. Relative paths are always specified
with respect to the location of the nhdr file.
The datafile
argument is not processed by write.nrrd
so it is
up to the caller to decide whether a relative or absolute path will be
used.
For write.nrrd.header.for.file
if outfile
is not specified
then the nhdr file will be placed next to the original image stack and the
datafile
field will therefore just be basename(infile)
. If
outfile is specified explicitly, then datafile
will be set to the
full path in the infile
argument. Therefore if you wish to specify
outfile
, you must set the current working directory (using
setwd
) to the location in which outfile
will be written to
ensure that the path to the datafile is correct. A future TODO would add
the ability to convert an absolute datafile path to a relative one (by
finding the common path between datafile and nhdr folders).
For write.nrrd
, arguments enc
, dtype
,
and endian
along with the dimensions of the input (x
) will
override the corresponding NRRD header fields from any supplied
header
argument. See
http://teem.sourceforge.net/nrrd/format.html for details of the NRRD
fields.
Other nrrd:
is.nrrd()
,
nrrd.voxdims()
,
read.nrrd()