fixspace.cpp

一ＯＣＲ的相关资料。．希望对研究ＯＣＲ的朋友有所帮助．

/******************************************************************
 * File:        fixspace.cpp  (Formerly fixspace.c)
 * Description: Implements a pass over the page res, exploring the alternative
 *					spacing possibilities, trying to use context to improve the
          word spacing
* Author:		Phil Cheatle
* Created:		Thu Oct 21 11:38:43 BST 1993
*
* (C) Copyright 1993, Hewlett-Packard Ltd.
** Licensed under the Apache License, Version 2.0 (the "License");
** you may not use this file except in compliance with the License.
** You may obtain a copy of the License at
** http://www.apache.org/licenses/LICENSE-2.0
** Unless required by applicable law or agreed to in writing, software
** distributed under the License is distributed on an "AS IS" BASIS,
** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
** See the License for the specific language governing permissions and
** limitations under the License.
*
**********************************************************************/

#include "mfcpch.h"
#include          <ctype.h>
#include          "reject.h"
#include          "statistc.h"
#include          "genblob.h"
#include          "control.h"
#include          "fixspace.h"
#include          "tessvars.h"
#include          "tessbox.h"
#include          "secname.h"

#define EXTERN

EXTERN BOOL_VAR (fixsp_check_for_fp_noise_space, TRUE,
"Try turning noise to space in fixed pitch");
EXTERN BOOL_VAR (fixsp_fp_eval, TRUE, "Use alternate evaluation for fp");
EXTERN BOOL_VAR (fixsp_noise_score_fixing, TRUE, "More sophisticated?");
EXTERN INT_VAR (fixsp_non_noise_limit, 1,
"How many non-noise blbs either side?");
EXTERN double_VAR (fixsp_small_outlines_size, 0.28, "Small if lt xht x this");

EXTERN BOOL_VAR (fixsp_ignore_punct, TRUE, "In uniform spacing calc");
EXTERN BOOL_VAR (fixsp_numeric_fix, TRUE, "Try to deal with numeric punct");
EXTERN BOOL_VAR (fixsp_prefer_joined_1s, TRUE, "Arbitrary boost");
EXTERN BOOL_VAR (tessedit_test_uniform_wd_spacing, FALSE,
"Limit context word spacing");
EXTERN BOOL_VAR (tessedit_prefer_joined_punct, FALSE,
"Reward punctation joins");
EXTERN INT_VAR (fixsp_done_mode, 1, "What constitues done for spacing");
EXTERN INT_VAR (debug_fix_space_level, 0, "Contextual fixspace debug");
EXTERN STRING_VAR (numeric_punctuation, ".,",
"Punct. chs expected WITHIN numbers");

#define PERFECT_WERDS   999
#define MAXSPACING      128      /*max expected spacing in pix */

/*************************************************************************
 * fix_fuzzy_spaces()
 * Walk over the page finding sequences of words joined by fuzzy spaces. Extract
 * them as a sublist, process the sublist to find the optimal arrangement of
 * spaces then replace the sublist in the ROW_RES.
 *************************************************************************/

void fix_fuzzy_spaces(                               //find fuzzy words
                      volatile ETEXT_DESC *monitor,  //progress monitor
                      INT32 word_count,              //count of words in doc
                      PAGE_RES *page_res) {
  BLOCK_RES_IT block_res_it;     //iterators
  ROW_RES_IT row_res_it;
  WERD_RES_IT word_res_it_from;
  WERD_RES_IT word_res_it_to;
  WERD_RES *word_res;
  WERD_RES_LIST fuzzy_space_words;
  INT16 new_length;
  BOOL8 prevent_null_wd_fixsp;   //DONT process blobless wds
  INT32 word_index;              //current word

  block_res_it.set_to_list (&page_res->block_res_list);
  word_index = 0;
  for (block_res_it.mark_cycle_pt ();
  !block_res_it.cycled_list (); block_res_it.forward ()) {
    row_res_it.set_to_list (&block_res_it.data ()->row_res_list);
    for (row_res_it.mark_cycle_pt ();
    !row_res_it.cycled_list (); row_res_it.forward ()) {
      word_res_it_from.set_to_list (&row_res_it.data ()->word_res_list);
      while (!word_res_it_from.at_last ()) {
        word_res = word_res_it_from.data ();
        while (!word_res_it_from.at_last () &&
          !(word_res->combination ||
          word_res_it_from.data_relative (1)->
          word->flag (W_FUZZY_NON) ||
          word_res_it_from.data_relative (1)->
        word->flag (W_FUZZY_SP))) {
          fix_sp_fp_word (word_res_it_from, row_res_it.data ()->row);
          word_res = word_res_it_from.forward ();
          word_index++;
          if (monitor != NULL) {
            monitor->ocr_alive = TRUE;
            monitor->progress = 90 + 5 * word_index / word_count;
          }
        }

        if (!word_res_it_from.at_last ()) {
          word_res_it_to = word_res_it_from;
          prevent_null_wd_fixsp =
            word_res->word->gblob_list ()->empty ();
          if (check_debug_pt (word_res, 60))
            debug_fix_space_level.set_value (10);
          word_res_it_to.forward ();
          word_index++;
          if (monitor != NULL) {
            monitor->ocr_alive = TRUE;
            monitor->progress = 90 + 5 * word_index / word_count;
          }
          while (!word_res_it_to.at_last () &&
            (word_res_it_to.data_relative (1)->
            word->flag (W_FUZZY_NON) ||
            word_res_it_to.data_relative (1)->
          word->flag (W_FUZZY_SP))) {
            if (check_debug_pt (word_res, 60))
              debug_fix_space_level.set_value (10);
            if (word_res->word->gblob_list ()->empty ())
              prevent_null_wd_fixsp = TRUE;
            word_res = word_res_it_to.forward ();
          }
          if (check_debug_pt (word_res, 60))
            debug_fix_space_level.set_value (10);
          if (word_res->word->gblob_list ()->empty ())
            prevent_null_wd_fixsp = TRUE;
          if (prevent_null_wd_fixsp)
            word_res_it_from = word_res_it_to;
          else {
            fuzzy_space_words.assign_to_sublist (&word_res_it_from,
              &word_res_it_to);
            fix_fuzzy_space_list (fuzzy_space_words,
              row_res_it.data ()->row);
            new_length = fuzzy_space_words.length ();
            word_res_it_from.add_list_before (&fuzzy_space_words);
            for (;
              (!word_res_it_from.at_last () &&
            (new_length > 0)); new_length--) {
              word_res_it_from.forward ();
            }
          }
          if (test_pt)
            debug_fix_space_level.set_value (0);
        }
        fix_sp_fp_word (word_res_it_from, row_res_it.data ()->row);
        //Last word in row
      }
    }
  }
}


void fix_fuzzy_space_list(  //space explorer
                          WERD_RES_LIST &best_perm,
                          ROW *row) {
  INT16 best_score;
  WERD_RES_LIST current_perm;
  INT16 current_score;
  BOOL8 improved = FALSE;

                                 //default score
  best_score = eval_word_spacing (best_perm);

  dump_words (best_perm, best_score, 1, improved);

  if (best_score != PERFECT_WERDS)
    initialise_search(best_perm, current_perm); 

  while ((best_score != PERFECT_WERDS) && !current_perm.empty ()) {
    match_current_words(current_perm, row); 
    current_score = eval_word_spacing (current_perm);
    dump_words (current_perm, current_score, 2, improved);
    if (current_score > best_score) {
      best_perm.clear ();
      best_perm.deep_copy (&current_perm);
      best_score = current_score;
      improved = TRUE;
    }
    if (current_score < PERFECT_WERDS)
      transform_to_next_perm(current_perm); 
  }
  dump_words (best_perm, best_score, 3, improved);
}


void initialise_search(WERD_RES_LIST &src_list, WERD_RES_LIST &new_list) { 
  WERD_RES_IT src_it(&src_list); 
  WERD_RES_IT new_it(&new_list); 
  WERD_RES *src_wd;
  WERD_RES *new_wd;

  for (src_it.mark_cycle_pt (); !src_it.cycled_list (); src_it.forward ()) {
    src_wd = src_it.data ();
    if (!src_wd->combination) {
      new_wd = new WERD_RES (*src_wd);
      new_wd->combination = FALSE;
      new_wd->part_of_combo = FALSE;
      new_it.add_after_then_move (new_wd);
    }
  }
}


void match_current_words(WERD_RES_LIST &words, ROW *row) { 
  WERD_RES_IT word_it(&words); 
  WERD_RES *word;

  for (word_it.mark_cycle_pt (); !word_it.cycled_list (); word_it.forward ()) {
    word = word_it.data ();
    if ((!word->part_of_combo) && (word->outword == NULL))
      classify_word_pass2(word, row); 
  }
}


/*************************************************************************
 * eval_word_spacing()
 * The basic measure is the number of characters in contextually confirmed
 * words. (I.e the word is done)
 * If all words are contextually confirmed the evaluation is deemed perfect.
 *
 * Some fiddles are done to handle "1"s as these are VERY frequent causes of
 * fuzzy spaces. The problem with the basic measure is that "561 63" would score
 * the same as "56163", though given our knowledge that the space is fuzzy, and
 * that there is a "1" next to the fuzzy space, we need to ensure that "56163"
 * is prefered.
 *
 * The solution is to NOT COUNT the score of any word which has a digit at one
 * end and a "1Il" as the character the other side of the space.
 *
 * Conversly, any character next to a "1" within a word is counted as a positive
 * score. Thus "561 63" would score 4 (3 chars in a numeric word plus 1 side of
 * the "1" joined).  "56163" would score 7 - all chars in a numeric word + 2
 * sides of a "1" joined.
 *
 * The joined 1 rule is applied to any word REGARDLESS of contextual
 * confirmation.  Thus "PS7a71 3/7a" scores 1 (neither word is contexutally
 * confirmed. The only score is from the joined 1. "PS7a713/7a" scores 2.
 *
 *************************************************************************/
INT16 eval_word_spacing(WERD_RES_LIST &word_res_list) { 
  WERD_RES_IT word_res_it(&word_res_list); 
  INT16 total_score = 0;
  INT16 word_count = 0;
  INT16 done_word_count = 0;
  INT16 word_len;
  INT16 i;
  WERD_RES *word;                //current word
  INT16 prev_word_score = 0;
  BOOL8 prev_word_done = FALSE;
  BOOL8 prev_char_1 = FALSE;     //prev ch a "1/I/l"?
  BOOL8 prev_char_digit = FALSE; //prev ch 2..9 or 0
  BOOL8 current_char_1 = FALSE;
  BOOL8 current_word_ok_so_far;
  STRING punct_chars = "!\"`',.:;";
  BOOL8 prev_char_punct = FALSE;
  BOOL8 current_char_punct = FALSE;
  BOOL8 word_done = FALSE;

  do {
    word = word_res_it.data ();
    word_done = fixspace_thinks_word_done (word);
    word_count++;
    if (word->tess_failed) {
      total_score += prev_word_score;
      if (prev_word_done)
        done_word_count++;
      prev_word_score = 0;
      prev_char_1 = FALSE;
      prev_char_digit = FALSE;
      prev_word_done = FALSE;
    }
    else {
      /*
        Can we add the prev word score and potentially count this word?
        Yes IF it didnt end in a 1 when the first char of this word is a digit
          AND it didnt end in a digit when the first char of this word is a 1
      */
      word_len = word->reject_map.length ();
      current_word_ok_so_far = FALSE;
      if (!((prev_char_1 &&
        digit_or_numeric_punct (word,
        word->best_choice->string ()[0])) ||
        (prev_char_digit &&
        ((word_done &&
        (word->best_choice->string ()[0] == '1')) ||
        (!word_done &&
        STRING (conflict_set_I_l_1).contains (word->best_choice->
      string ()[0])))))) {
        total_score += prev_word_score;
        if (prev_word_done)
          done_word_count++;
        current_word_ok_so_far = word_done;
      }

      if ((current_word_ok_so_far) &&
        (!tessedit_test_uniform_wd_spacing ||
        ((word->best_choice->permuter () == NUMBER_PERM) ||
      uniformly_spaced (word)))) {
        prev_word_done = TRUE;
        prev_word_score = word_len;
      }
      else {
        prev_word_done = FALSE;
        prev_word_score = 0;
      }

      if (fixsp_prefer_joined_1s) {
        /* Add 1 to total score for every joined 1 regardless of context and rejtn */

        for (i = 0, prev_char_1 = FALSE; i < word_len; i++) {
          current_char_1 = word->best_choice->string ()[i] == '1';
          if (prev_char_1 || (current_char_1 && (i > 0)))
            total_score++;
          prev_char_1 = current_char_1;
        }
      }

      /* Add 1 to total score for every joined punctuation regardless of context
        and rejtn */
      if (tessedit_prefer_joined_punct) {