Nicolai Grossherr – X Y Z

ogr – one go ocr – automation script for ocr

—

, ,

https://github.com/grossherr/ogr

#!/bin/bash

# ogr - one go ocr
# Git: https://git.nicolaigrossherr.xyz 
# Doc: https://nicolaigrossherr.xyz

# variables
SCRIPTNAME=${0##*/}
SCRIPTVERSION="0.3.3"


function ParameterSetup() {
	VariableDeclarer

	while [[ $# -gt 0 ]]
	do
		parameter="$1"
		case $parameter in
			-a|--action)
			action_variable=${2}
			shift
			shift
			;;
			-S|Scanner)
			action_variable="Scanner"
			shift
			;;
			-T)
			action_variable="Test"
			shift
			;;
    	-i|--input)
	    input_variable="${2}"
  	  shift 
    	shift 
	    ;;
  	  -o|--output)
    	output_variable="${2}"
		  shift 
	    shift 
  	  ;;
    	--tdir)
	    tmp_dir="$2"
  	  shift 
 		 	shift 
			;;
    	--default)
   		DEFAULT=YES
    	shift 
    	;;
    	*)    
    	additional_parameter+=("$1") 
    	shift
    	;;
		esac
	done
	set -- "${additional_parameter[@]}"

	VariableSetter

#printf "\n"
#printf "$action_variable\n"
#printf "$input_variable\n"
#printf "$output_variable\n"
#printf "\n"
}


function VariableDeclarer() {
	declare -g action_variable 
	declare -g input_variable
	declare -g output_variable

	declare -a additional_parameter

	declare -g is_automation
	declare -g last_func

	declare -g cores

	declare -g verbose_status
	declare -g measure_time

	declare -g tmp_dir
	declare -A tmp_dirs
	declare -g tmp_info

	declare -g input_dir
	declare -g output_dir

	declare -g input_file
	declare -g output_file

	declare -g file_prefix
	declare -g file_number
}

function VariableSetter() {
	## directories
	if [ -z $tmp_dir ]; then
		tmp_suffix=$( date +%s )
		tmp_dir="./tmp_${SCRIPTNAME}_${tmp_suffix}"
	fi
	tmp_info="$tmp_dir/info"
	input_dir="."
	output_dir="."

	## files
	input_file=""
	output_file=""
	file_prefix="page"
	file_number="%03d"
	image_file_ext="tif"

	## settings
	dpi="600"

	### programs
	uext="pbm"
	ulay="single"

	## multi-threading
	cores=$( nproc )

	## process
	last_func=""

	## verbose
	verbose_status=0
	measure_time=0
}

#if [ ! -z "${input_variable}" ]; then
#	input_variable_file=${input_variable##*/}
#	input_variable_dir=${input_variable%/*}
#	if [ ! -z "${input_variable_dir}" ]; then
#		input_dir=$input_variable_dir
#	fi
#	if [ ! -z "${input_variable_file}" ]; then
#		input_file=$input_variable_file
#	fi
#	if [ "${input_file}" = "${input_dir}" ]; then
#		input_dir="./"
#	fi
#	printf "$input_variable_dir\n"
#	printf "$input_variable_file\n"
#fi

#if [ ! -z "${output_variable}" ]; then
#	output_variable_file=${output_variable##*/}
#	output_variable_dir=${output_variable%/*}
#	if [ ! -z "${output_variable_dir}" ]; then
#		output_dir=$output_variable_dir
#	fi
#	if [ ! -z "${output_variable_file}" ]; then
#		output_file=$output_variable_file
#	fi
#	if [ "${output_file}" = "${output_dir}" ]; then
#		output_dir="./"
#	fi
#	printf "$output_variable_dir\n"
#	printf "$output_variable_file\n"
#fi

function Test() {
	echo "$action_variable"
	echo "$1"
	echo "$2"
	echo "$3"
	echo "$4"
}


## helper
function ScriptHeader() {
	printf "$SCRIPTNAME $SCRIPTVERSION\n\n"
}

function PrintStep() {
#	printf "${#FUNCNAME[@]}"
	fn_depth=$((${#FUNCNAME[@]} - 3))

	if [ $fn_depth -gt 0 ] && [ $verbose_status -gt 1 ]
	then
		printf ' %.0s' $(seq 1 $fn_depth)
	fi
	
	if [ $fn_depth -lt 2 ] && [ -z $input_dir ] || [ $verbose_status -gt 1 ]
	then
		if [ ! -z $2 ]; then printf "$2"; fi
		if [ -z "$1" ]; then
			printf "${FUNCNAME[ 1 ]}"
		else
			printf "$1"
		fi
		printf "\n"
	fi
}

function ExecTime() {
	if [ -z $1 ]
	then
		return 1
	fi

	declare -A -g exec_time
	exec_start="start_time"
	exec_end="end_time"
	exec_res="execution_time"

	if [ $1 = "start" ]
	then
		if [ -z $2 ]
		then
			exec_time[ $exec_start ]="$(date +%s)"
		elif [ ! -z $2 ]
		then
			arg1_exec_start=$2_$exec_start
			exec_time[ "$arg1_exec_start" ]="$(date +%s)"
		fi
	fi

	if [ $1 = "end" ] 
	then
		if [ -z $2 ] && [ ${exec_time[ $exec_start ]} ]
		then
			exec_time[ $exec_end ]="$(date +%s)"
			exec_time[ $exec_res ]="$(echo "${exec_time[ $exec_end ]} - ${exec_time[ $exec_start ]}" | bc)"
			printf "${exec_time[ "$exec_res" ]}"
		elif [ ! -z $2 ] && [ ${exec_time[ "$arg1_exec_start" ]} ]
		then
			arg1_exec_start=$2_$exec_start
			arg1_exec_end=$2_$exec_end
			arg1_exec_res=$2_$exec_res
			exec_time[ "$arg1_exec_end" ]="$(date +%s)"
			exec_time[ "$arg1_exec_res" ]="$(echo "${exec_time[ "$arg1_exec_end" ]} - ${exec_time[ "$arg1_exec_start" ]}" | bc)"
			printf "${exec_time[ "$arg1_exec_res" ]}"
		fi
		if [ -z $3 ] 
		then 
			printf "\n"
		fi
	fi
}

function TimeAnalyser() {
	if [ $verbose_status -gt 1 ] || [ $measure_time -eq 1 ]
	then
		if [ ! -z $3 ]
		then
			printf "$3"
		fi
		ExecTime $1 $2
	fi
}

function InOutDirHandler() {
#	for l in $( seq 0 $(( ${#FUNCNAME[@]} - 1 )) ); do
#		printf "${FUNCNAME[ $l ]} fnl\n"
#	done
#	printf "$LINENO: $last_func\n"
#	printf "$LINENO: ${tmp_dirs[ $last_func ]}\n"

	func_call_level="${#FUNCNAME[@]}"

	calling_func=${FUNCNAME[ 1 ]}
#	printf "$LINENO: $calling_func\n"

	if [ -z $input_dir ] && [ ${tmp_dirs[ $last_func ]} ]; then
		input_dir=${tmp_dirs[ $last_func ]}
	fi

	tmp_dirs[ $calling_func ]=$tmp_dir/${calling_func,,}
#	printf "$LINENO: ${tmp_dirs[ $calling_func ]}\n"

	if [ -z $output_dir ] && [ $func_call_level = "4" ]; then
		output_dir="${tmp_dirs[ $calling_func ]}"
	fi

	if [ ! -d $output_dir ]; then
		mkdir -p $output_dir
	fi


}

function LastFunction() {
	if [ -z $1 ]; then
		level="2"
	else
		level=$1
	fi
	last_func=${FUNCNAME[ $level ]}
#	printf "Dbg $FUNCNAME: $last_func\n"
}

function UnsetInOutDir() {
	input_dir=""
	output_dir=""
}

function SilentPost() {
	LastFunction
	UnsetInOutDir
}



# setup
function Setup() {
	PrintStep

	DeleteTmpDirs
	UnsetInOutDir
}
## delete temporary directories
function DeleteTmpDirs() {
	PrintStep "Deleting $tmp_dir"
	
	rm -Rf $tmp_dir
}
## create temporary directories
function CreateTmpDirs() {
	PrintStep "Creating $tmp_dir"

	for directory in $directories 
	do
		if [ ! -d "$directory" ]
		then
			mkdir "$directory"
		fi
	done
}

function File() {
	Setup "$@"

	PrintStep 

	TimeAnalyser start

	Files
	Rotate
	Trim
	Scantailor
	Deskew
	Unpaper
	Resize
	OCR
	Pdf

	TimeAnalyser end
}
function Files() {
	PrintStep

	TimeAnalyser start $FUNCNAME

	InOutDirHandler

	FileHandler

	SilentPost
	TimeAnalyser end $FUNCNAME
}
function FileHandler() {
	declare -a -g file_types
	file_types=(image/jpg application/pdf image/png image/tiff)
	mime=$( file --mime-type -b "${input_variable}" )

	printf "$LINENO: $input_variable\n"
	printf "$LINENO: $output_variable\n"
	printf "$LINENO: $input_dir\n"
	printf "$LINENO: $output_dir\n"

	if [[ " ${file_types[@]} " =~ " ${mime} " ]]
	then
		mkdir -p "${output_dir}"
		mime_ext=${mime##*/}
		if [ $mime_ext = "tiff" ]; then
			pages=$(identify "${input_variable}" | wc -l )
			if [ $pages -gt 1 ]
			then
				convert -scene 1 -type bilevel -density $dpi "${input_variable}" "${output_dir}/${file_prefix}${file_number}.${image_file_ext}"
			else
				cp "${1}" "${output_dir}/${output_file}"
			fi
		fi
	else
		printf "Not supported\n"
	fi
}


# script
function Scanner() {
	Setup "$@"

	PrintStep 

	TimeAnalyser start

	Scan
	Rotate
	Trim
	Scantailor
	Deskew
	Unpaper
	Resize
	OCR
	Pdf

	TimeAnalyser end
}

function ScanSingle() {
	PrintStep 

	TimeAnalyser start

	Scan

	TimeAnalyser end
}


# scanner
function Scan() {
	PrintStep

	TimeAnalyser start $FUNCNAME

	InOutDirHandler

	scan_file_prefix="$file_prefix"
	scan_file_number="$file_number"
	scan_device="brother3:net1;dev0"
#	scan_device="brother3:bus4;dev1"
	scan_color_mode="Black & White"
#	scan_color_mode="True Gray"
#	scan_source="Automatic Document Feeder(left aligned)"
	scan_source="Automatic Document Feeder(left aligned,Duplex)"
	scan_resolution="$dpi"
	scan_format="tiff"
	scan_file_extension=$scan_format
	scan_directory="$output_dir"
	scan_file_format="/$scan_file_prefix$scan_file_number.$scan_file_extension"
	scan_file_path="$scan_directory$scan_file_format"

	if [ $verbose_status -gt 2 ]
	then
		scanimage -B \
			--device-name "$scan_device" \
			--mode "$scan_color_mode" \
			--source "$scan_source" \
			--resolution "$scan_resolution" \
			--format=$scan_format \
			--batch=$scan_file_path \
			--progress \
			--verbose
	else
		scanimage -B \
			--device-name "$scan_device" \
			--mode "$scan_color_mode" \
			--source "$scan_source" \
			--resolution "$scan_resolution" \
			--format=$scan_format \
			--batch=$scan_file_path \
			--brightness -25 \
			--progress >/dev/null 2>&1
	fi

	SilentPost
	TimeAnalyser end $FUNCNAME
}

# rotate
function Rotate() {
	PrintStep

	TimeAnalyser start ${FUNCNAME[ 0 ]}

	InOutDirHandler

	AnalysingRotation
	RotateAuto

	SilentPost

	TimeAnalyser end ${FUNCNAME[ 0 ]}
}
function AnalysingRotation() {
	PrintStep

	TimeAnalyser start ${FUNCNAME[ 0 ]}

	local files=$( find $input_dir -maxdepth 1 -type f -name "*" )
	if [ ! -d $tmp_info ]; then
		mkdir -p $tmp_info
	fi
	local threads="$cores"

	for image in $files; do 
		file_from_path="${image##*/}"
		file_wo_ext="${file_from_path%.*}"
		((process=process%threads)); ((process++==0)) && wait
		if [ $verbose_status -gt 2 ]
		then
		OMP_THREAD_LIMIT=1 \
			tesseract --dpi $dpi \
			-c tessedit_page_number=0 \
			--psm 0 \
			$image \
			- \
			| awk '/Rotate/ {print $2}' > $tmp_info/${file_wo_ext}_rotation &
		else
		OMP_THREAD_LIMIT=1 \
			tesseract --dpi $(( $dpi / 2 )) \
			-c tessedit_page_number=0 \
			--psm 0 \
			$image \
			- \
			/dev/null 2>&1 \
			| awk '/Rotate/ {print $2}' > $tmp_info/${file_wo_ext}_rotation &
		fi
	done
	wait

	TimeAnalyser end ${FUNCNAME[ 0 ]} "›"
}
## rotate auto
function RotateAuto() {
	PrintStep

	TimeAnalyser start ${FUNCNAME[ 0 ]}

	if [ $verbose_status -gt 2 ]
	then
		verbose="-verbose"
	else 
		verbose=""
	fi

	local files=$( find $input_dir -maxdepth 1 -type f -name "*" )
	local threads="$cores"

	for image in $files; do 
		file_from_path="${image##*/}"
		file_wo_ext="${file_from_path%.*}"
		rotation=$( cat "$tmp_info/${file_wo_ext}_rotation" )
		((process=process%threads)); ((process++==0)) && wait
		convert $image \
			$verbose \
			-rotate \
			$rotation \
			$output_dir/${image##*/} &
	done 
	wait

	TimeAnalyser end ${FUNCNAME[ 0 ]} "›"
}
## rotate page
#function RotatePage() {}
## rotate backside
#function RotateBackside() {}

# trim
function Trim() {
	PrintStep

	TimeAnalyser start $FUNCNAME

	InOutDirHandler

	TrimAuto

	SilentPost

	TimeAnalyser end $FUNCNAME
}
# trim auto
function TrimAuto() {
	PrintStep

	if [ $verbose_status -gt 2 ]
	then
		verbose="-verbose"
	else 
		verbose=""
	fi

	local files=$( find $input_dir -maxdepth 1 -type f -name "*" )
	local threads="$cores"

	for image in $files; do
		((process=process%threads)); ((process++==0)) && wait
		convert $image \
			$verbose \
			-trim \
			+repage \
			$output_dir/${image##*/} &
	done
	wait
}

# scantailor
function Scantailor() {
	PrintStep

	TimeAnalyser start $FUNCNAME

	InOutDirHandler

	ScantailorCli

	SilentPost

	TimeAnalyser end $FUNCNAME
}
## scantailor-cli
function ScantailorCli() {
	PrintStep

	if [ $verbose_status -gt 2 ]
	then
		verbose="--verbose"
	else
		verbose=""
	fi

	local files=$( find $input_dir -maxdepth 1 -type f -name "*" )
	local threads="$cores"

	st_color_mode="color_grayscale"#black_and_white

	for image in $files; do
		((process=process%threads)); ((process++==0)) && wait
    scantailor-cli --start-filter=1 \
			--end-filter=6 \
			--normalize-illumination \
			--white-margins \
			--alignment-vertical=top \
			--alignment-horizontal=center \
			--margin=10 \
			--despeckle=aggressive \
			--content-detection=aggressive \
			--deskew=auto \
			--threshold=-5 \
			--dpi=$dpi \
			--output-dpi=$dpi \
			--color-mode=$st_color_mode \
			--layout=0 \
			$image \
			$output_dir &
	done
	wait
}

# deskew
function Deskew() {
	PrintStep

	TimeAnalyser start $FUNCNAME

	InOutDirHandler

	DeskewAuto

	SilentPost

	TimeAnalyser end $FUNCNAME
}
## deskew auto
function DeskewAuto() {
	PrintStep

	local files=$( find $input_dir -maxdepth 1 -type f -name "*" )
	local threads="$cores"

	for image in $files; do
		((process=process%threads)); ((process++==0)) && wait
		if [ $verbose_status -gt 2 ]
		then
			deskew -q lanczos \
				-b FFFFFFFF \
				-a 25 \
				-o $output_dir/${image##*/} \
				$image &
		else
			deskew -q lanczos \
				-b FFFFFFFF \
				-a 25 \
				-o $output_dir/${image##*/} \
				$image >/dev/null 2>/dev/null &
		fi
	done
	wait
}

# unpaper
function Unpaper() {
	PrintStep

	TimeAnalyser start $FUNCNAME

	InOutDirHandler

	UnpaperAuto

	SilentPost

	TimeAnalyser end $FUNCNAME
}
## unpaper auto
function UnpaperAuto() {
	PrintStep

	local files=$( find $input_dir -maxdepth 1 -type f -name "*" )
	local threads="$cores"

	for image in $files; do
		file_from_path="${image##*/}"
		fup="${file_from_path%.*}.$uext"
		((process=process%threads)); ((process++==0)) && wait
		if [ $verbose_status -gt 2 ]
		then
			unpaper --dpi $dpi \
				-t $uext \
				--verbose \
				--overwrite \
				$image \
				$output_dir/$fup &
		else
			unpaper --dpi $dpi \
				-t $uext \
				--overwrite \
				--no-deskew \
				$image \
				$output_dir/$fup \
				>/dev/null 2>&1 &
		fi
	done
	wait
}

# resize
function Resize() {
	PrintStep

	TimeAnalyser start $FUNCNAME

	InOutDirHandler

	ResizeAuto

	SilentPost

	TimeAnalyser end $FUNCNAME
}
## resize auto
function ResizeAuto() {
	PrintStep 

	if [ $verbose_status -gt 2 ]
	then
		verbose="-verbose"
	else
		verbose=""
	fi

	local files=$( find $input_dir -maxdepth 1 -type f -name "*" )
	local threads="$cores"

	for image in $files; do
		((process=process%threads)); ((process++==0)) && wait
		convert $image \
			$verbose \
			-type Bilevel \
			-compress Fax \
			-quality 9 \
			-resize 3496x \
			$output_dir/${image##*/} &
	done
	wait
}

# OCR
function OCR() {
	PrintStep

	TimeAnalyser start $FUNCNAME

	InOutDirHandler

	OCRsingle

	SilentPost

	TimeAnalyser end $FUNCNAME
}

## ocr single
function OCRsingle() {
	PrintStep

	local files=$( find $input_dir -maxdepth 1 -type f -name "*" )
	#local threads=$(( $cores / 4 ))
	local threads="$cores"

	for image in $files; do
		local file_from_path="${image##*/}"
		local file_wo_ext="${file_from_path%.*}"
		((process=process%threads)); ((process++==0)) && wait
		if [ $verbose_status -gt 2 ]
		then
			OMP_THREAD_LIMIT=1 \
				tesseract --dpi $(( $dpi / 2 )) \
				-l deu \
				--psm 12 \
				$image \
				$output_dir/$file_wo_ext \
				pdf txt &
		else
			OMP_THREAD_LIMIT=1 \
				tesseract --dpi $(( $dpi / 2 )) \
				-l deu \
				--psm 12 \
				$image \
				$output_dir/$file_wo_ext \
				pdf txt >/dev/null 2>&1 &
		fi
	done
	wait

}

## ocr auto
function OCRmulti() {
	PrintStep

	printf "$input_variable\n"
	printf "$output_variable\n"

	tesseract --dpi $dpi \
		-l deu \
		--psm 12 \
		${input_variable} \
		${output_variable} \
		pdf
}


function Pdf() {
	PrintStep

	TimeAnalyser start $FUNCNAME

	InOutDirHandler

	MergePdf

	SilentPost

	TimeAnalyser end $FUNCNAME
}
function MergePdf() {
	PrintStep

	gs -sDEVICE=pdfwrite \
		-r$(( $dpi / 4 )) \
		-sPAPERSIZE=a4 \
		-dPDFSETTINGS=/ebook \
		-dCompatibilityLevel=1.4 \
		-dNOPAUSE \
		-dQUIET \
		-dBATCH \
		-o ${output_dir}/result.pdf \
		${input_dir}/*.pdf
}

# Output
function MergeMpTif() {
	printf "\n› Merging\n"

	tiffcp $tpre/* $tmp_dir/merge.tif
}

function_dump=$( declare -f $1 >/dev/null 2>&1 )
function_exists="$?"
#echo "$function_exists"

#if declare -f $1 > /dev/null 
if [ $function_exists -eq 0 ]
then
	function_call=$1
	shift
	ParameterSetup "$@"
	"$function_call" "$@"
else
	ParameterSetup "$@"
#	echo "$action_variable"
	if declare -f $action_variable > /dev/null
	then 
		"$action_variable" "$@"
	elif [ ! -z $1 ]
	then
		printf "'$1' does not exist\n" >&2
	elif [ ! -z $action_variable ]
	then
		printf "'$action_variable' does not exist\n" >&2
	else
		printf "Parameter needed, see --help\n" >&2
	fi
	exit 1
fi

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *