fixspace.cpp

一ＯＣＲ的相关资料。．希望对研究ＯＣＲ的朋友有所帮助．

    tprintf ("FP fixspace working on \"%s\"\n",
      word_res->best_choice->string ().string ());
  }
  #endif
  gblob_sort_list ((PBLOB_LIST *) word_res->word->rej_cblob_list (), FALSE);
  sub_word_list_it.add_after_stay_put (word_res_it.extract ());
  fix_noisy_space_list(sub_word_list, row); 
  new_length = sub_word_list.length ();
  word_res_it.add_list_before (&sub_word_list);
  for (; (!word_res_it.at_last () && (new_length > 1)); new_length--) {
    word_res_it.forward ();
  }
}


void fix_noisy_space_list(WERD_RES_LIST &best_perm, ROW *row) { 
  INT16 best_score;
  WERD_RES_IT best_perm_it(&best_perm); 
  WERD_RES_LIST current_perm;
  WERD_RES_IT current_perm_it(&current_perm); 
  WERD_RES *old_word_res;
  WERD_RES *new_word_res;
  INT16 current_score;
  BOOL8 improved = FALSE;

                                 //default score
  best_score = fp_eval_word_spacing (best_perm);

  dump_words (best_perm, best_score, 1, improved);

  new_word_res = new WERD_RES;
  old_word_res = best_perm_it.data ();
                                 //Kludge to force deep copy
  old_word_res->combination = TRUE;
  *new_word_res = *old_word_res; //deep copy
                                 //Undo kludge
  old_word_res->combination = FALSE;
                                 //Undo kludge
  new_word_res->combination = FALSE;
  current_perm_it.add_to_end (new_word_res);

  break_noisiest_blob_word(current_perm); 

  while ((best_score != PERFECT_WERDS) && !current_perm.empty ()) {
    match_current_words(current_perm, row); 
    current_score = fp_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)
      break_noisiest_blob_word(current_perm); 
  }
  dump_words (best_perm, best_score, 3, improved);
}


/*************************************************************************
 * break_noisiest_blob_word()
 * Find the word with the blob which looks like the worst noise.
 * Break the word into two, deleting the noise blob.
 *************************************************************************/
void break_noisiest_blob_word(WERD_RES_LIST &words) { 
  WERD_RES_IT word_it(&words); 
  WERD_RES_IT worst_word_it;
  float worst_noise_score = 9999;
  int worst_blob_index = -1;     //noisiest blb of noisiest wd
  int blob_index;                //of wds noisiest blb
  float noise_score;             //of wds noisiest blb
  WERD_RES *word_res;
  C_BLOB_IT blob_it;
  C_BLOB_IT rej_cblob_it;
  C_BLOB_LIST new_blob_list;
  C_BLOB_IT new_blob_it;
  C_BLOB_IT new_rej_cblob_it;
  WERD *new_word;
  INT16 start_of_noise_blob;
  INT16 i;

  for (word_it.mark_cycle_pt (); !word_it.cycled_list (); word_it.forward ()) {
    blob_index = worst_noise_blob (word_it.data (), &noise_score);
    if ((blob_index > -1) && (worst_noise_score > noise_score)) {
      worst_noise_score = noise_score;
      worst_blob_index = blob_index;
      worst_word_it = word_it;
    }
  }
  if (worst_blob_index < 0) {
    words.clear ();              //signal termination
    return;
  }

  /* Now split the worst_word_it */

  word_res = worst_word_it.data ();

  /* Move blobs before noise blob to a new bloblist */

  new_blob_it.set_to_list (&new_blob_list);
  blob_it.set_to_list (word_res->word->cblob_list ());
  for (i = 0; i < worst_blob_index; i++, blob_it.forward ()) {
    new_blob_it.add_after_then_move (blob_it.extract ());
  }
  start_of_noise_blob = blob_it.data ()->bounding_box ().left ();
  delete blob_it.extract ();     //throw out noise blb

  new_word = new WERD (&new_blob_list, word_res->word);
  new_word->set_flag (W_EOL, FALSE);
  word_res->word->set_flag (W_BOL, FALSE);
  word_res->word->set_blanks (1);//After break

  new_rej_cblob_it.set_to_list (new_word->rej_cblob_list ());
  rej_cblob_it.set_to_list (word_res->word->rej_cblob_list ());
  for (;
    (!rej_cblob_it.empty () &&
    (rej_cblob_it.data ()->bounding_box ().left () <
  start_of_noise_blob)); rej_cblob_it.forward ()) {
    new_rej_cblob_it.add_after_then_move (rej_cblob_it.extract ());
  }

  worst_word_it.add_before_then_move (new WERD_RES (new_word));

  word_res->done = FALSE;
  if (word_res->outword != NULL) {
    delete word_res->outword;
    delete word_res->best_choice;
    delete word_res->raw_choice;
    word_res->outword = NULL;
    word_res->best_choice = NULL;
    word_res->raw_choice = NULL;
  }
}


INT16 worst_noise_blob(WERD_RES *word_res, float *worst_noise_score) { 
  PBLOB_IT blob_it;
  INT16 blob_count;
  float noise_score[512];
  int i;
  int min_noise_blob;            //1st contender
  int max_noise_blob;            //last contender
  int non_noise_count;
  int worst_noise_blob;          //Worst blob
  float small_limit = bln_x_height * fixsp_small_outlines_size;
  float non_noise_limit = bln_x_height * 0.8;

  blob_it.set_to_list (word_res->outword->blob_list ());
  //normalised
  blob_count = blob_it.length ();
  ASSERT_HOST (blob_count <= 512);
  if (blob_count < 5)
    return -1;                   //too short to split
  /* Get the noise scores for all blobs */

  #ifndef SECURE_NAMES
  if (debug_fix_space_level > 5)
    tprintf ("FP fixspace Noise metrics for \"%s\": ",
      word_res->best_choice->string ().string ());
  #endif

  for (i = 0; i < blob_count; i++, blob_it.forward ()) {
    if (word_res->reject_map[i].accepted ())
      noise_score[i] = non_noise_limit;
    else
      noise_score[i] = blob_noise_score (blob_it.data ());

    if (debug_fix_space_level > 5)
      tprintf ("%1.1f ", noise_score[i]);
  }
  if (debug_fix_space_level > 5)
    tprintf ("\n");

  /* Now find the worst one which is far enough away from the end of the word */

  non_noise_count = 0;
  for (i = 0;
  (i < blob_count) && (non_noise_count < fixsp_non_noise_limit); i++) {
    if (noise_score[i] >= non_noise_limit)
      non_noise_count++;
  }
  if (non_noise_count < fixsp_non_noise_limit)
    return -1;
  min_noise_blob = i;

  non_noise_count = 0;
  for (i = blob_count - 1;
  (i >= 0) && (non_noise_count < fixsp_non_noise_limit); i--) {
    if (noise_score[i] >= non_noise_limit)
      non_noise_count++;
  }
  if (non_noise_count < fixsp_non_noise_limit)
    return -1;
  max_noise_blob = i;

  if (min_noise_blob > max_noise_blob)
    return -1;

  *worst_noise_score = small_limit;
  worst_noise_blob = -1;
  for (i = min_noise_blob; i <= max_noise_blob; i++) {
    if (noise_score[i] < *worst_noise_score) {
      worst_noise_blob = i;
      *worst_noise_score = noise_score[i];
    }
  }
  return worst_noise_blob;
}


float blob_noise_score(PBLOB *blob) { 
  OUTLINE_IT outline_it;
  BOX box;                       //BB of outline
  INT16 outline_count = 0;
  INT16 max_dimension;
  INT16 largest_outline_dimension = 0;

  outline_it.set_to_list (blob->out_list ());
  for (outline_it.mark_cycle_pt ();
  !outline_it.cycled_list (); outline_it.forward ()) {
    outline_count++;
    box = outline_it.data ()->bounding_box ();
    if (box.height () > box.width ())
      max_dimension = box.height ();
    else
      max_dimension = box.width ();

    if (largest_outline_dimension < max_dimension)
      largest_outline_dimension = max_dimension;
  }

  if (fixsp_noise_score_fixing) {
    if (outline_count > 5)
                                 //penalise LOTS of blobs
      largest_outline_dimension *= 2;

    box = blob->bounding_box ();

    if ((box.bottom () > bln_baseline_offset * 4) ||
      (box.top () < bln_baseline_offset / 2))
                                 //Lax blob is if high or low
      largest_outline_dimension /= 2;
  }
  return largest_outline_dimension;
}


void fixspace_dbg(WERD_RES *word) { 
  BOX box = word->word->bounding_box ();
  BOOL8 show_map_detail = FALSE;
  INT16 i;

  box.print ();
  #ifndef SECURE_NAMES
  tprintf (" \"%s\" ", word->best_choice->string ().string ());
  tprintf ("Blob count: %d (word); %d/%d (outword)\n",
    word->word->gblob_list ()->length (),
    word->outword->gblob_list ()->length (),
    word->outword->rej_blob_list ()->length ());
  word->reject_map.print (debug_fp);
  tprintf ("\n");
  if (show_map_detail) {
    tprintf ("\"%s\"\n", word->best_choice->string ().string ());
    for (i = 0; word->best_choice->string ()[i] != '\0'; i++) {
      tprintf ("**** \"%c\" ****\n", word->best_choice->string ()[i]);
      word->reject_map[i].full_print (debug_fp);
    }
  }

  tprintf ("Tess Accepted: %s\n", word->tess_accepted ? "TRUE" : "FALSE");
  tprintf ("Done flag: %s\n\n", word->done ? "TRUE" : "FALSE");
  #endif
}


/*************************************************************************
 * fp_eval_word_spacing()
 * Evaluation function for fixed pitch word lists.
 *
 * Basically, count the number of "nice" characters - those which are in tess
 * acceptable words or in dict words and are not rejected.
 * Penalise any potential noise chars
 *************************************************************************/

INT16 fp_eval_word_spacing(WERD_RES_LIST &word_res_list) { 
  WERD_RES_IT word_it(&word_res_list); 
  WERD_RES *word;
  PBLOB_IT blob_it;
  INT16 word_length;
  INT16 score = 0;
  INT16 i;
  const char *chs;
  float small_limit = bln_x_height * fixsp_small_outlines_size;

  if (!fixsp_fp_eval)
    return (eval_word_spacing (word_res_list));

  for (word_it.mark_cycle_pt (); !word_it.cycled_list (); word_it.forward ()) {
    word = word_it.data ();
    word_length = word->reject_map.length ();
    chs = word->best_choice->string ().string ();
    if ((word->done ||
      word->tess_accepted) ||
      (word->best_choice->permuter () == SYSTEM_DAWG_PERM) ||
      (word->best_choice->permuter () == FREQ_DAWG_PERM) ||
      (word->best_choice->permuter () == USER_DAWG_PERM) ||
    (safe_dict_word (chs) > 0)) {
      blob_it.set_to_list (word->outword->blob_list ());
      for (i = 0; i < word_length; i++, blob_it.forward ()) {
        if ((chs[i] == ' ') ||
          (blob_noise_score (blob_it.data ()) < small_limit))
          score -= 1;            //penalise possibly erroneous non-space

        else if (word->reject_map[i].accepted ())
          score++;
      }
    }
  }
  if (score < 0)
    score = 0;
  return score;
}