This is a simple pipeline to reduce the memory requirements
for Vesuvius scroll sections. It builds on the great work
by james darby to mask out the background material of the
individual slices.

For visualising data and testing ideas I have found
that 8-bit data is very capable. The pipeline takes
images and clips them between given pixel thresholds.
I have used 18000 and 65535 to increase the contrast
of the scroll fibres (see figure). Then saved the
resultant images as 8-bit PNGs. This results in the
whole of scroll 1 fitting in approximately 270 Gb of
storage.

The pipeline for generating a mask is shown below and
has been tested on scrolls 1, 0332 and 1667.

1. Place the main.py and support.py in the folder
   containing the original images.
2. Open the main file in your editor of choice and
   choose the desired settings (see suggested_settings.txt
   for scrolls). If you wish to delete the original files
   as you go, set REMOVE_ORIGINAL equal to True.
3. Run the python program from the terminal using:
   python main.py

This is a relatively involved script for generating a
histogram of pixel values for a given region in an image.

1. Open script and update the image source.
2. Update the bounds to set a region with an x_min, y_min,
   x_max, y_max format.
3. Run the python program for the terminal using:
   python pixel_analysis.py
4. Review the output region.png and histogram.png files.

Python 3.x
Python libraries (os, sys, cv2, numpy, matplotlib, collections)

To clip and mask scroll 0332 we first need to characterise
the pixel values. This can be achieved using the pixel_analysis
script. This shows that the center bright core is mostly pixel
values of >55000 and that the air is typically around 15000-20000.
Thus potential clipping points are 18000 and 55000, to help
increase the contrast of the papyrus sheets.

Following this, a flood point is required that will be used to
generate a mask. This is a point on the image that is always within
the case. Once all the settings are prepared it is useful to run
a test with each 1000th slice to check how the pipeline is performing.

The results clearly show that throughout the scroll the pipeline
seems to be masking the case successfully.