Double Mask

[Tyler W. Davis]

Try applying a second mask to the image processes.

1. Take the standard deviation of the image sequence (as was previously thought for determining boundaries), scale the values to 0--255, and threshold to create a mask of pixels with high variability; the thought is that the mesh doesn't move, so it may masked by this approach. This mask is used for each image in the sequence.

2. Use the color mask that is currently created for each individual image in the sequence

[Tyler W. Davis]

Addresses #2 - created color mask function

→ <<cset fc59da78510f>>

[Tyler W. Davis]

Addresses #2 - early implement of positional mask

→ <<cset ada611bcdb53>>

[Tyler W. Davis]

Positional mask seems not to work due to the plant being in both images.

TODO: try imaging a mesh tower without a plant to create a background image that could be subtracted from images with both mesh and plant

[Tyler W. Davis]

Try this:

img = imopen(my_file)
cmax = numpy.max(img, 2)
cmin = numpy.min(img, 2)    # looks a lot like blue channel
delta = (cmax - cmin)       # highlights mesh over roots
cmin_bin = threshold(cmin)
delta_bin = threshold(delta)
mask = (cmin_bin != delta_bin) & cmin_bin 

[Tyler W. Davis]

v0.8.0. Addresses #1 - new delta mask does a better job of removing mesh from some sequences. Addresses #2 - delta mask replaces positional mask, which performed poorly.

→ <<cset 8216a3692437>>

[Tyler W. Davis]

Two imaging masks are currently implemented (color and delta) and both can be used simultaneously, resulting in a third mask defined by their union.