# Last edited on 2025-08-21 19:41:31 by stolfi
# 080 analysis of the multispectral images
  
  This note describes the application of several image processing
  and analysis methods applied to the multispectral images of the VMS
  captured on 2014-08-19.
  
SETUP

  ln -s ../.. work
  ln -s work/error_funcs.py
  ln -s work/process_funcs.py
  ln -s work/image_funcs.py
  ln -s ${HOME}/lib/argparser.py

  ln -s ../../../MultiSpectral
  ln -s MultiSpectral MS
  ln -s MultiSpectral/davis/bands_table_funcs.py 
  
SOURCE OF THE MULTI-SPECTRAL IMAGES

  Lisa Fagin Davis, formerly Beinecke staff, posted to GoggleDrive
  the higher resoltion multi-spectral scans of some pages
  captured at the Beinecke Library on 2014-08-19.
  
  Downloaded and converted the TIFF scans for some pages, and converted
  them to 16-bit grayscale PNG. See "Multispectral/davis/00-Notebook.txt".
  
  One big question is whether the samples in those TIFF files are gamma
  encoded or not. A color patch scale was included in the images, but we
  do not know their albedos at the wavelengths in question.
  
  The raw images are in 
  
    "MS/davis/{page}/{band}/full.tif"
    "MS/davis/{page}/{band}/full.pgm"

  For each page, the processing extracted small rectangles with selected details:
  
    "MS/davis/{page}/{band}/clips/C1.pgm"
    "MS/davis/{page}/{band}/clips/C2.pgm"

ALTERNATE SOURCE

  User @oshfdk of the Voynich Ninja website set up a webpage displaying
  the Beinecke multispectral images if pages f70r1 (Aries Darke) and
  f71r (Aries Light). Some of those images were downloaded to
  "Multispectral/oshfdk/"
    
CREATING COMPOSITES BY HAND
    
  Created a composite of IR, green, and UV clips:
  
    ./make_false_color_clips_by_hand.sh 001r1
    ./make_false_color_clips_by_hand.sh 008r1
  
CONVERSION TO FLOAT IMAGES

  Wrote a script to convert the TIFF images to PNG and to numpy 
  arrays:

    convert_clips_to_numpy.py 008r1 C1 C2
    
  This script requires a file "MS/davis/{page}/bands.txt" listing the
  color bands and their attributes: type of illumination (0 = frontal, 1
  and 2 = grazing, 3 = from behind), the dominant wavelength, and the
  max useful sample value (for scaling to [0_1]).
    
SPECTRAL ANALYSIS

  Manually created black (ignore) and white (take) masks in
  "pages/008r1/clips/C1/masks":
  
    wr/mask.png  the written area excluding the leaves.
    ch/mask.png  only the glyphs with a bit of background around them.

    bg/mask.png  a sample of background (bare parch).
    gb/mask.png  a sample of background with ghost leaf from f8v.

    di/mask.png  a sample of the darker glyph strokes (Rt2?).
    mi/mask.png  a sample of the prevalent glyph strokes (Rt1?)
    li/mask.png  a sample of the faintest glyph strokes (original?)

    gp/mask.png  a sample of the green paint from the leaf.
    yp/mask.png  a sample of the light yellow (yel) paint from the stem.
    
  The masks "ch.png" and "tx.png" include the outline of the plant's stem
  but exclude all the painted areas.

  Wrote three programs to analyze the spectra of pixls in thise
  areas
  
    analyze_clip_mask_pixel_spectra.py 008r1 C1 ch V0
    create_eigen_images.py 008r1 V0 C1 wr
  
# ----------------------------------------------------------------------
OLD STUFF

  Displaying the maxval-adjusted monochrme images:
    
    for page in 008r1 ; do
      set -e
      for band in ` cd deriv/davis/${page} && ls -d {MB,TX}*_F ` ; do
        display -title '%d' deriv/davis/${page}/${band}/C1.png 
      done
      set +e
    done