Hacking at the Voynich manuscript - Side notes
086 Enhancing text and drawings covered by green paint

Last edited on 2025-10-23 17:39:57 by stolfi

INTRODUCTION


  This note is about enhancing text and drawings that were covered by green
  paint, by doing color processing on the Beinecke 2014 RGB images.

SETUP

    ln -s ../work

    ln -s work/error_funcs.gawk

    ln -s work/error_funcs.py
    ln -s work/process_funcs.py

    ln -s work/vms_linear_gray_image_funcs.py
    ln -s work/vms_color_image_funcs.py
    ln -s work/vms_multispectral_image_funcs.py
    
    ln -s ${HOME}/lib/argparser.py
    ln -s ${HOME}/ttf

EXTRACTING THE SOURCE IMAGES

  The source images will be placed in the folder "clips/{page}/{clip}/data".  
  
  A good size for the pools at bottom of Bio pages seems to be 2040x1024
  as it covers the whole big green pool at the bottom of pages f79v,
  f82r, and f82v. The green pool at the SW corner of f79r will be
  narrower.

    "022v1/leaf2w/data/clip.png"  f22v  800x420+824+1158
    "035r1/ochre1/data/clip.png"  f35r  1300x1300+1158+1000
    "079r1/poolsw/data/clip.png"  f79r  800x1024+63+2385
    "079v1/poolbt/data/clip.png"  f79v  2048x1024+647+2545
    "082r1/poolbt/data/clip.png"  f82r  2048x1024+70+2519
    "082v1/poolbt/data/clip.png"  f82v  2048x1024+690+2411
    
  Also created clips of the corresponding areas on the other side of the folio,
  horzly flipped, to check for possible bleedthroughs.
  
    "022v1/leaf2w/data/pilc.png"  f22r 800x420+1172+1105  (f22v[x,y] ~ f22r[2796-x,y-53]
    "035r1/ochre1/data/pilc.png"  f35v 1304x1310+439+1009 scale to 1300x1300
    "079r1/poolsw/data/pilc.png"  f79v 800x1024+1940+2360 (f79r[x,y] ~ f79v[2803-x,y-23])  
    
MASKS

  Created in folder "clips/{page}/{clip}/masks" some masks that select pixels with specific
  colors, such as
  
    "parch.png"   Blank parchment with no color bleeding.
    "bleed.png"   Blank parchment with green color bleeding from the other side.
    "green.png"   Areas painted green not suspected of covering inked stuff.
    "dkink.png"   Dark text and figure penned ink (suspected Rt2/Rt3).
    
  etc.

PROCESSING THE IMAGES

  Wrote a python program to separate pixels into classes using Bayes's formula, given
  samples of pixels from each class:
  
    bayesian_ink_sep.py
    
  See the Makefile for calls.