#!/bin/bash
#
# divide_vert [-i] image  multi_image
#
# Divide an image horizontally into rows of images consisting of alternating
# images of variable colors (interesting parts) and images of a single solid
# color (seperating gaps) that when appended together vertically will
# re-produce the original image.
#
# Options...
#    -i   Don't output (delete) the uninteresting 'gaps' between the rows.
#
# Examples...
#
# Expand the row spacing of some text by 4 pixels (must be a multiple of two)
#
#  convert -size 50x caption:'This is a word wrapped line of text' miff:- |\
#    divide_vert - miff:- | convert - -splice 0x2 -append  gap_expansion.png
#
# This will remove all gaps between lines of text...
#
#  convert -size 50x caption:'This is a word wrapped line of text' miff:- |\
#    divide_vert -i - miff:- | convert - -append  gaps-removed.png
#
# Replace all gaps by a fixed amount (5 pixels),
# though this is not a good idea for normal text images.
#
#  convert -size 50x caption:'This is a word wrapped line of text' miff:- |\
#    divide_vert -i - miff:- |\
#      convert - -splice 0x5 -append \
#              -gravity south -splice 0x5  gaps_replaced.png
#
# If you want to split on gaps of a specific color you can use -splice or
# -border to add a single column of that color along the left edge.
#
# FUTURE:
#   * Expand to allow fuzzy matching of the colors in each row and gap.
#     This is needed for scanned and JPEG images.
#   * Allow dividing the image into 'columns' instead of 'rows'.
#   * Specify a minimum 'gap' size, to allow 'paragraph/word' seperation
#   * Optionally divide the 'gap' equally bettween each 'interesting' row That
#     is every image contains 'interesting' parts, surrounded by some 'blank'
#     of 'gap' spacing.
#   * Save images with 'virtual (page) offsets' to preserve their location.
#     The result will be a collection of 'layer' images rather than 'simple'
#     rows.  such 'layer' images will then allow you to separate an image
#     first vertically by 'row', then horizontally by 'column' producing
#     a exploded layed image of either individual characters or words.
#     It will also let you do automated  "GIF animation splitting".
#
###
#
# Anthony Thyssen    25 June 2007     A.Thyssen_AT_griffith.edu.au
#
PROGNAME=`type $0 | awk '{print $3}'`  # search for executable on path
PROGDIR=`dirname $PROGNAME`            # extract directory of program
PROGNAME=`basename $PROGNAME`          # base name of program
Usage() {                              # output the script comments as docs
  echo >&2 "$PROGNAME:" "$@"
  sed >&2 -n '/^###/q; /^#/!q; s/^#//; s/^ //; 3s/^/Usage: /; 2,$ p' \
          "$PROGDIR/$PROGNAME"
  exit 10;
}

while [  $# -gt 0 ]; do
  case "$1" in
  --help|--doc*) Usage ;;
  -i)   IGNORE_GAPS=true ;;    # do not output blank gap images.
  --) shift; break ;;    # end of user options
  -)  break ;;           # STDIN end of user options
  -*) Usage "Unknown option \"$1\"" ;;
  *)  break ;;           # end of user options
  esac
  shift   # next option
done

input="$1"  # Read from this IM input file format, may be a pipe
output="$2" # Write to this IM input file format, may be a pipe


# Handle the many temporary files needed.
tmpdir=/tmp/$PROGNAME.$$
mkdir $tmpdir
if [ $? -ne 0 ]; then
  echo >&2 "$PROGNAME: Failed to create directory \"$tmpdir\" -- ABORTING"
  exit 1
fi
trap "rm -rf $tmpdir; exit 0" 0
trap "rm -rf $tmpdir; exit 1" 1 2 3 15

# Read input image, once only as it may be pipelines, and save a MPC copy of it.
# Return the height of that image.
height=`convert "$input" -format '%[fx:u.h-1]' -identify $tmpdir/original.mpc`

count=0   # the number of divisions found in the file

# Divide image into individual rows and
# Get the number of colors, and first color from each row!
convert $tmpdir/original.mpc -crop 0x1 +repage \
        -format '%k %[pixel:s]' info:$tmpdir/image_data

exec <$tmpdir/image_data     # read from this data as STDIN
read top_cnum top_color      # get the color count, and first pixel color
top=0                        # the top row of current area being divided

# Divide the image basied on the number of colors (or change in gap color)
for x in `seq 1 $height`; do
  read cnum color
  if [ $top_cnum -eq 1 -a \( $cnum -ne 1 -o "$top_color" != "$color" \) ] ||
     [ $top_cnum -ne 1 -a $cnum -eq 1 ]; then
    # We found the end of a variable image or single color 'blank' gap
    if [ $top_cnum -ne 1 -o ! "$IGNORE_GAPS" ]; then
      # Save this division from the original image
      h=`expr $x - $top`   # height of region
      file=`printf "$tmpdir/divisions_%06d.mpc" $count`  # save the division
      convert $tmpdir/original.mpc -crop 0x$h+0+$top +repage $file
    fi
    # note the information for the next area to be divided
    top=$x   top_cnum=$cnum  top_color="$color"
    count=`expr $count + 1`  # we found this many divisions.
  fi
done

if [ $count -eq 0 ]; then
  # No divisions were found, so just output image (via convert again)
  if [ $top_cnum -ne 1 -o ! "$IGNORE_GAPS" ]; then
    convert $tmpdir/original.mpc "$output"
  else
    # image is blank, BUT we were asked to not output blank images!
    # This is imposible, so ouput the 'null' error image
    convert null: "$output"
  fi
else
  # Extract the final split (if wanted)
  if [ $top_cnum -ne 1 -o ! "$IGNORE_GAPS" ]; then
    # save the last division that is present in the original image
    h=`expr $height + 1 - $top`   # height of region
    file=`printf "$tmpdir/divisions_%06d.mpc" $count`
    convert $tmpdir/original.mpc -crop 0x$h+0+$top +repage $file
  fi
  #count=`expr $count + 1`
  #echo >&2 "split_horiz: Found $count divisions"

  # output the parts that were saved, if any!
  if  ls -d $tmpdir/divisions_*.mpc >/dev/null 2>/dev/null;  then
    # Output all the divisions as a posible multi-image file
    convert $tmpdir/divisions_*.mpc "$output"
  else
    # No divisions were found! presumably we are ignoring gaps!
    # This should be imposible, so ouput the 'null' error image
    convert null: "$output"
  fi
fi