When down-scaling an image and labelling it the volumes of the particles of the scaled image are not (bin^3) times smaller than those of the unscaled image.
I have a tiff image of a single particle which is already labelled (label 1109, image size in px z=111, y=260, x=234 and pixel size 0.0122486 mm). I import the tiff file (Image = tifffile.imread( "OnePArticle-ID1109.tif" )
) and run Vol = ltk.Volumes( Image )
and then save it with numpy.savetxt('Vol_unscaled.tsv',Vol)
. Then I scale down the Image using Image_scaled = scipy.ndimage.zoom( Image, 0.5, order=1 )
, run Vol_scaled = ltk.Volumes( Image_scaled )
and save it with numpy.savetxt('Vol_scaled.tsv',Vol_scaled)
. The tsv of the Vol_unscaled gives me a particle with ID 1109 and a volume of 744268 vx (which has already been validated) and the rest of the ID's are zero. The tsv of the Vol_scaled gives me a particle with ID 1109 and a volume of 90543 vx (90543*(2^3) = 724344), a number of other smaller particles with a total volume of 4987. In this example the volume of the particle is underestimated and a lot of other "particles" are created which can be easily filtered out in the images where the rest of the particles are significantly larger. In another example I did the same (with the size of the image being such that can be divided by 2), the volumes were overestimated. I have attached the image of the single labelled particle and the tsv files.ParticleVolume.zip