/util.py

"""Utility functions for processing images for delivery to Tesseract"""



import os

import re



_add_dot_txt_flag = False





def image_to_scratch(im, scratch_image_name):

	"""Saves image in memory to scratch file.  .bmp format will be read correctly by Tesseract"""

##	if im.mode=='RGBA':

##		im=im.convert('RGB')

##	try:

##		im.save(scratch_image_name, dpi=(200,200))

##	except:  ### Eventually this should catch only the specific im.save exception

##		im = im.convert('RGB')

##		im.save(scratch_image_name, dpi=(200,200))

	#im = im.convert('1')

	im.save(scratch_image_name, dpi=(200,200))



def	retrieve_text(scratch_text_name_root):

	if _add_dot_txt_flag:

		inf = file(scratch_text_name_root + '.txt')

	else:

		inf = file(scratch_text_name_root)

	text = inf.read().strip()

	inf.close()

	return text



# Eventually this more thorough class should be used:



##class OCR_character:

##	"""Object exposing internals of Tesseract result for particular characters

##	(See documentation of EANYCODE_CHAR

##	http://tesseract-ocr.googlecode.com/svn&cs_f=trunk/ccutil/ocrclass.h

##	for detailed explanations)

##	self.letter - OCRed letter guess

##	self.char_code - Character code of letter

##	self.x_bounds - (left bound, right bound)

##	self.y_bounds - (top bound, bottom bound)

##	self.font_index - Index of character's font

##	self.confidence - 0 (low conf) to 100 (high)

##	self.point_size - Estimated size of font (units unclear)

##	self.formatting - Bit flags for formatting and layout information

##	"""

##	def __init__(self, line):

##		data = line.split(' ')

##		self.letter = data[0]

##		self.char_code = int(data[1], 16)

##		self.x_bounds = (data[2], data[4])

##		self.y_bounds = (data[5], data[3])

##		self.font_index = data[6]

##		self.confidence = data[7]

##		self.point_size = data[8]

##		self.formatting = data[9]

##	def __str__(self):

##		return self.letter



# This simple class is used for now:



class OCR_character:

	"""Object exposing internals of Tesseract result for particular characters

	(See documentation of EANYCODE_CHAR

	http://tesseract-ocr.googlecode.com/svn&cs_f=trunk/ccutil/ocrclass.h

	for detailed explanations)

	self.letter - OCRed letter guess

	self.x_bounds - (left bound, right bound)

	self.y_bounds - (top bound, bottom bound)

	"""

	def __init__(self, line):

		parse_re = re.compile(r'^(.).*\((.+),(.+)\).*\((.+),(.+)\)') # Match example 'T[54]->[54](35,115)->(56,90)'

		data = parse_re.findall(line)[0]

		self.letter = data[0]

		self.x_bounds = (int(data[1]), int(data[3]))

		self.y_bounds = (int(data[2]), int(data[4]))

	



class OCR_result(str):

	"""Parsed results of call to Tesseract; subclass of 'str'.

	self OCR string.

	self.internals is array (aligned with self.text) of OCR_letter

		internal data objects (for characters) or None (for whitespace,

		since Tesseract provides no internal data for whitespace characters)."""

	def __new__(self, text):

		raw_letters = []

		internals = []

		data = text.split('\n')

		i = 0

		while i<len(data):

			line = data[i].strip()

			if line=='<nl>':  # New line

				raw_letters.append('\n')

				internals.append(None)

				i += 1

			elif line=='<para>':  # End of input

				break

			elif line=='':  # Space character

				raw_letters.append(' ')

				internals.append(None)

			else:

				character = OCR_character(line)

				raw_letters.append(character.letter)

				internals.append(character)

			i += 1

		self = str.__new__(self, "".join(raw_letters))

		self.internals = internals

		return self

			

	

def retrieve_result(scratch_text_name_root):

	text = retrieve_text(scratch_text_name_root)

	return OCR_result(text)



def perform_cleanup(scratch_image_name, scratch_text_name_root):

	"""Clean up temporary files from disk"""

	if _add_dot_txt_flag:

		scratch_text_name = scratch_text_name_root + '.txt'

	else:

		scratch_text_name = scratch_text_name_root

	for name in (scratch_image_name, scratch_text_name, "tesseract.log"):

		try:

			os.remove(name)

		except OSError:

			pass