GitLab

In the_"not_parallelized" version I write the new function that
will be parallelized and checj it works

Int the "yes_parallelized" version, the
multiprocessing.map method is used.

some prints went back

AND

surface voxels file must be in the path_read input
(if computed, flag_SV = True) it will be saved in
path_input instead of path_save

an image with connected pores (obtained using
the FIJI's pluggin connected regions over the
void's phase of a segmentation)
and computes
the volume of each reagion
the min and max X,Y,Z coordinate
the lenght of the pore in each direction (of the
 inscribing box)
a bool saying if the pore is connected to the top
of the sample (top = when z = 0)

specific surfaces where the FM/voids surface
is divided by
1) FM volume
2) Air Volume
3) FM+Air Volume
4) Air + Grv + FM volume

in the past commit

ne marchait pas bien.
Celle là oui!

set by the user OR
computed by the program

Remember
When you set manually the max size of pores to
computehistograms you can do comparisons
between different Local Thickness maps

whole sample.

So the different zones of computation are
Whole sample
Deposit Layer and Gravel Layer mixed togueter ***
Dep Layer
Grv Layer

*** The criterion we defined for the deposit
and gravel layer makes that the sum of the two regions
is smaller than the whole sample

That is why I added the new line for the whole sample
*** O

counter in FIJI.
But we won't use that pluggin anymore to compute
connected pores.

Instead, we are using the connected regions pluggin

using the erosion_dilation method for pore
size distribution (PSD) anymore

enhacement in FIJI is already well parallelized,
so it is much faster.

Though in FIJI, YOU MUST BE VERY CAREFUL WHEN YOU DO
!!SLICE!! OR !!STACK!! CONTRAST ENHANCEMENT
NORMALIZATION because sometimes stack based
normalization will not work... I don't know why

Concatenation of raw image (when you have to
concatenate scans that had to be splitted in two
or more parts during at the laboratory) is not as
simple as one may think to do it based on computations

The best way is to do it manually, and to do it
choosing a slice that will not present noise.

I say noise, becasie whe you do several scans of
an object, at the beggining and at the end of the
scan you may find an artifact like a shadow
sweeping the image. Hopefully, this shadow
disappears at the center of the image.

In consequence, when you do a manual concatenation,
pick the most inner image you can !

Don't forget there is a script, that concatenates
all parts once you have found the indexes of the
slices where you want to concatenate

was enhanced as for article 2 (spatial representativity
of Xr-CT) we decided to analyze electrical conductivity
and not anymore the relative saturation content.

Yes it sounds weird...

FIJI will compute the connected regions (CR) but
you may have RAM limitations. So you will have
to split your original image in different parts.

This split will connect this parts as long as you
have left one slice of "recouvrement" (covering zone)
when you splitted the original image.

The programm deals fine with all the possibilities
of connections due to the splitting of the image
and will reassign new labels for every pore.

perform computations in cylinders smaller
than total diameter

Adding a "rw" parameter, which means the the fraction
of the radius that will be taken into account
to compute the profiles.
In the relabel function, pixels ouside a cilinder
of radius r = len(image)*"rw" will be labeled
with a value = 1 -> pixels outside the cylinder

Not very important:
A new parameter sample_name: to create output
files with shorter names, not like.

eliminating unuseful lists

the new one computes everything, global volume fractions, global volumes, and depths
the 3 of them for the deposit layer and the gravel layer once the limits have been established by the vertical volume fraction profiles

all these changes were writen to follow in the message file the evolution of the growinf procedure