minc.ray.trace {RMINC} | R Documentation |
Call ray_trace to get an image of a rendered slice
minc.ray.trace("filename.mnc") minc.ray.trace(volume, output="slice.rgb", size=c(400,400), slice=list(pos=0, wv="w", axis="z"), threshold=NULL, colourmap="-spectral", background=NULL, background.threshold=NULL, background.colourmap="-gray", display=TRUE)
volume |
The filename of a volume to render. |
output |
The output filename. |
size |
A vector of two elements specifying the output size |
slice |
A list of three elements, pos being the slice number, wv whether the specification is in voxel or world space, and the axis. |
threshold |
A vector of two elements containing the threshold. If NULL, the full range of the volume will be used. |
colourmap |
The colourmap to be used by ray_trace. |
background |
An optional filename of a background volume. Used, for example, to render statistical results on top of background anatomy. |
background.threshold |
Threshold to use for the background volume. If NULL the whole range will be used. |
background.colourmap |
The colourmap argument to be passed to ray_trace for the background image. |
display |
Boolean argument which determines whether display (from ImageMagick) will be called on the output. |
This function provides an interface to the ray_trace command written by David MacDonald. As such it needs both ray_trace and make_slice to be on the path upon startup of R, and the bicpl library has to be compiled with image output enabled.
Behaviour of minc.ray.trace varies depending on whether a background image is specified. If background=NULL, then the specified slice is rendered using the supplied (or automatically determined) threshold argument. If there is a background image, then the slice from the input volume is rendered semi-transparently on top of the background.
Note that cropping in ray_trace is on by default, so the output image size will not necessarily be the same as the size argument to minc.ray.trace.
output |
The filename of the output image is returned. |
minc.apply, minc.write.volume, minc.model
## Not run: # get a file that could be used by glim_image gf <- as.data.frame(read.table("filename.glim")) # get a t-test at every voxel t.stats <- minc.model(gf$V1, gf$V2, "t-test") # and write to file minc.write.volume("t-test.mnc", gf$V1[1], t.stats) # create a pretty picture to include in the next Nature or Science # article minc.ray.trace("t-test.mnc", output="pretty.rgb", threshold=c(2.5,6), background="anatomy_image.mnc") ## End(Not run)