C++ WERD::flag方法代码示例

INT16 resegment_box(  //
                    ROW *row,
                    BOX box,
                    char *ch,
                    INT16 block_id,
                    INT16 row_id,
                    INT16 boxfile_lineno,
                    INT16 boxfile_charno) {
  WERD_IT word_it;
  WERD *word;
  WERD *new_word = NULL;
  BOOL8 polyg = false;
  PBLOB_IT blob_it;
  PBLOB_IT new_blob_it;
  PBLOB *blob;
  PBLOB *new_blob;
  OUTLINE_IT outline_it;
  OUTLINE_LIST dummy;  // Just to initialize new_outline_it.
  OUTLINE_IT new_outline_it = &dummy;
  OUTLINE *outline;
  BOX new_word_box;
  float word_x_centre;
  float baseline;
  INT16 error_count = 0;         //number of chars lost

  word_it.set_to_list (row->word_list ());
  for (word_it.mark_cycle_pt (); !word_it.cycled_list (); word_it.forward ()) {
    word = word_it.data ();
    polyg = word->flag (W_POLYGON);
    if (word->bounding_box ().overlap (box)) {
      blob_it.set_to_list (word->gblob_list ());
      for (blob_it.mark_cycle_pt ();
      !blob_it.cycled_list (); blob_it.forward ()) {
        blob = blob_it.data ();
        if (gblob_bounding_box (blob, polyg).overlap (box)) {
          outline_it.set_to_list (gblob_out_list (blob, polyg));
          for (outline_it.mark_cycle_pt ();
          !outline_it.cycled_list (); outline_it.forward ()) {
            outline = outline_it.data ();
            if (goutline_bounding_box (outline, polyg).
            major_overlap (box)) {
              if (strlen (word->text ()) > 0) {
                if (error_count == 0) {
                  error_count = 1;
                  if (applybox_debug > 4)
                    report_failed_box (boxfile_lineno,
                      boxfile_charno,
                      box, ch,
                      "FAILURE! box overlaps blob in labelled word");
                }
                if (applybox_debug > 4)
                  tprintf
                    ("APPLY_BOXES: ALSO ignoring corrupted char blk:%d row:%d \"%s\"\n",
                    block_id, row_id,
                    word_it.data ()->text ());
                word_it.data ()->set_text ("");
                //UN label it
                error_count++;
              }

              if (error_count == 0) {
                if (new_word == NULL) {
                                 /* Make a new word with a single blob */
                  new_word = word->shallow_copy ();
                  new_word->set_text (ch);
                  if (polyg)
                    new_blob = new PBLOB;
                  else
                    new_blob = (PBLOB *) new C_BLOB;
                  new_blob_it.set_to_list (new_word->
                    gblob_list ());
                  new_blob_it.add_to_end (new_blob);
                  new_outline_it.
                    set_to_list (gblob_out_list
                    (new_blob, polyg));
                }
                new_outline_it.add_to_end (outline_it.
                  extract ());
                //move blob
              }
            }
          }
                                 //no outlines in blob
          if (outline_it.empty ())
                                 //so delete blob
            delete blob_it.extract ();
        }
      }
      if (blob_it.empty ())      //no blobs in word
                                 //so delete word
          delete word_it.extract ();
    }
  }
  if (error_count > 0)
    return error_count;

  if (new_word != NULL) {
    gblob_sort_list (new_word->gblob_list (), polyg);
    word_it.add_to_end (new_word);
    new_word_box = new_word->bounding_box ();
//.........这里部分代码省略.........

//.........这里部分代码省略.........
    tprintf ("Leftmost blob is at (%d,%d)\n",
      box_it.data ()->bounding_box ().left (),
      box_it.data ()->bounding_box ().bottom ());
    return nullptr;
  }
  ASSERT_HOST (!cell_it.empty () && !row->char_cells.singleton ());
  prev_chop_coord = cell_it.data ()->x ();
  word = nullptr;
  while (rep_left < cell_it.data ()->x ()) {
    word = add_repeated_word (&rep_it, rep_left, prev_chop_coord,
      blanks, row->fixed_pitch, &word_it);
  }
  cell_it.mark_cycle_pt ();
  if (prev_chop_coord >= cell_it.data ()->x ())
    cell_it.forward ();
  for (; !cell_it.cycled_list (); cell_it.forward ()) {
    chop_coord = cell_it.data ()->x ();
    while (!box_it.empty ()
    && box_it.data ()->bounding_box ().left () <= chop_coord) {
      if (box_it.data ()->bounding_box ().right () > prev_x)
        prev_x = box_it.data ()->bounding_box ().right ();
      split_to_blob (box_it.extract (), chop_coord,
        textord_fp_chop_error + 0.5f,
        &left_coutlines,
        &right_coutlines);
      box_it.forward ();
      while (!box_it.empty() && box_it.data()->cblob() == nullptr) {
        delete box_it.extract();
        box_it.forward();
      }
    }
    if (!right_coutlines.empty() && left_coutlines.empty())
      split_to_blob (nullptr, chop_coord,
        textord_fp_chop_error + 0.5f,
        &left_coutlines,
        &right_coutlines);
    if (!left_coutlines.empty()) {
      cblob_it.add_after_then_move(new C_BLOB(&left_coutlines));
    } else {
      if (rep_left < chop_coord) {
        if (rep_left > prev_chop_coord)
          new_blanks = (uint8_t) floor ((rep_left - prev_chop_coord)
            / row->fixed_pitch + 0.5);
        else
          new_blanks = 0;
      }
      else {
        if (chop_coord > prev_chop_coord)
          new_blanks = (uint8_t) floor ((chop_coord - prev_chop_coord)
            / row->fixed_pitch + 0.5);
        else
          new_blanks = 0;
      }
      if (!cblob_it.empty()) {
        if (blanks < 1 && word != nullptr && !word->flag (W_REP_CHAR))
          blanks = 1;
        word = new WERD (&cblobs, blanks, nullptr);
        cblob_it.set_to_list (&cblobs);
        word->set_flag (W_DONT_CHOP, TRUE);
        word_it.add_after_then_move (word);
        if (bol) {
          word->set_flag (W_BOL, TRUE);
          bol = false;
        }
        blanks = new_blanks;
      }
      else
        blanks += new_blanks;
      while (rep_left < chop_coord) {
        word = add_repeated_word (&rep_it, rep_left, prev_chop_coord,
          blanks, row->fixed_pitch, &word_it);
      }
    }
    if (prev_chop_coord < chop_coord)
      prev_chop_coord = chop_coord;
  }
  if (!cblob_it.empty()) {
    word = new WERD(&cblobs, blanks, nullptr);
    word->set_flag (W_DONT_CHOP, TRUE);
    word_it.add_after_then_move (word);
    if (bol)
      word->set_flag (W_BOL, TRUE);
  }
  ASSERT_HOST (word != nullptr);
  while (!rep_it.empty ()) {
    add_repeated_word (&rep_it, rep_left, prev_chop_coord,
      blanks, row->fixed_pitch, &word_it);
  }
                                 //at end of line
  word_it.data ()->set_flag (W_EOL, TRUE);
  if (prev_chop_coord > prev_x)
    prev_x = prev_chop_coord;
  xstarts[1] = prev_x + 1;
  real_row = new ROW (row, (int16_t) row->kern_size, (int16_t) row->space_size);
  word_it.set_to_list (real_row->word_list ());
                                 //put words in row
  word_it.add_list_after (&words);
  real_row->recalc_bounding_box ();
  return real_row;
}

ROW *find_row_of_box(                         //
                     BLOCK_LIST *block_list,  //real blocks
                     BOX box,                 //from boxfile
                     INT16 &block_id,
                     INT16 &row_id_to_process) {
  BLOCK_IT block_it(block_list);
  BLOCK *block;
  ROW_IT row_it;
  ROW *row;
  ROW *row_to_process = NULL;
  INT16 row_id;
  WERD_IT word_it;
  WERD *word;
  BOOL8 polyg;
  PBLOB_IT blob_it;
  PBLOB *blob;
  OUTLINE_IT outline_it;
  OUTLINE *outline;

  /*
    Find row to process - error if box REALLY overlaps more than one row. (I.e
    it overlaps blobs in the row - not just overlaps the bounding box of the
    whole row.)
  */

  block_id = 0;
  for (block_it.mark_cycle_pt ();
  !block_it.cycled_list (); block_it.forward ()) {
    block_id++;
    row_id = 0;
    block = block_it.data ();
    if (block->bounding_box ().overlap (box)) {
      row_it.set_to_list (block->row_list ());
      for (row_it.mark_cycle_pt ();
      !row_it.cycled_list (); row_it.forward ()) {
        row_id++;
        row = row_it.data ();
        if (row->bounding_box ().overlap (box)) {
          word_it.set_to_list (row->word_list ());
          for (word_it.mark_cycle_pt ();
          !word_it.cycled_list (); word_it.forward ()) {
            word = word_it.data ();
            polyg = word->flag (W_POLYGON);
            if (word->bounding_box ().overlap (box)) {
              blob_it.set_to_list (word->gblob_list ());
              for (blob_it.mark_cycle_pt ();
              !blob_it.cycled_list (); blob_it.forward ()) {
                blob = blob_it.data ();
                if (gblob_bounding_box (blob, polyg).
                overlap (box)) {
                  outline_it.
                    set_to_list (gblob_out_list
                    (blob, polyg));
                  for (outline_it.mark_cycle_pt ();
                    !outline_it.cycled_list ();
                  outline_it.forward ()) {
                    outline = outline_it.data ();
                    if (goutline_bounding_box
                    (outline, polyg).major_overlap (box)) {
                      if ((row_to_process == NULL) ||
                      (row_to_process == row)) {
                        row_to_process = row;
                        row_id_to_process = row_id;
                      }
                      else
                        /* RETURN ERROR Box overlaps blobs in more than one row  */
                        return NULL;
                    }
                  }
                }
              }
            }
          }
        }
      }
    }
  }
  return row_to_process;
}

void Textord::clean_noise_from_words(          //remove empties
                                     ROW *row  //row to clean
                                    ) {
  TBOX blob_box;                 //bounding box
  C_BLOB *blob;                  //current blob
  C_OUTLINE *outline;            //current outline
  WERD *word;                    //current word
  int32_t blob_size;             //biggest size
  int32_t trans_count;           //no of transitions
  int32_t trans_threshold;       //noise tolerance
  int32_t dot_count;             //small objects
  int32_t norm_count;            //normal objects
  int32_t dud_words;             //number discarded
  int32_t ok_words;              //number remaining
  int32_t word_index;            //current word
                                 //words of row
  WERD_IT word_it = row->word_list ();
  C_BLOB_IT blob_it;             //blob iterator
  C_OUTLINE_IT out_it;           //outline iterator

  ok_words = word_it.length ();
  if (ok_words == 0 || textord_no_rejects)
    return;
  // was it chucked
  std::vector<int8_t> word_dud(ok_words);
  dud_words = 0;
  ok_words = 0;
  word_index = 0;
  for (word_it.mark_cycle_pt (); !word_it.cycled_list (); word_it.forward ()) {
    word = word_it.data ();      //current word
    dot_count = 0;
    norm_count = 0;
                                 //blobs in word
    blob_it.set_to_list (word->cblob_list ());
    for (blob_it.mark_cycle_pt (); !blob_it.cycled_list ();
    blob_it.forward ()) {
      blob = blob_it.data ();
      if (!word->flag (W_DONT_CHOP)) {
                                 //get outlines
        out_it.set_to_list (blob->out_list ());
        for (out_it.mark_cycle_pt (); !out_it.cycled_list ();
        out_it.forward ()) {
          outline = out_it.data ();
          blob_box = outline->bounding_box ();
          blob_size =
            blob_box.width () >
            blob_box.height ()? blob_box.width () : blob_box.
            height();
          if (blob_size < textord_noise_sizelimit * row->x_height ())
            dot_count++;         //count smal outlines
          if (!outline->child ()->empty ()
            && blob_box.height () <
            (1 + textord_noise_syfract) * row->x_height ()
            && blob_box.height () >
            (1 - textord_noise_syfract) * row->x_height ()
            && blob_box.width () <
            (1 + textord_noise_sxfract) * row->x_height ()
            && blob_box.width () >
            (1 - textord_noise_sxfract) * row->x_height ())
            norm_count++;        //count smal outlines
        }
      }
      else
        norm_count++;
      blob_box = blob->bounding_box ();
      blob_size =
        blob_box.width () >
        blob_box.height ()? blob_box.width () : blob_box.height ();
      if (blob_size >= textord_noise_sizelimit * row->x_height ()
      && blob_size < row->x_height () * 2) {
        trans_threshold = blob_size / textord_noise_sizefraction;
        trans_count = blob->count_transitions (trans_threshold);
        if (trans_count < textord_noise_translimit)
          norm_count++;
      }
      else if (blob_box.height () > row->x_height () * 2
        && (!word_it.at_first () || !blob_it.at_first ()))
        dot_count += 2;
    }
    if (dot_count > 2 && !word->flag(W_REP_CHAR)) {
      if (dot_count > norm_count * textord_noise_normratio * 2)
        word_dud[word_index] = 2;
      else if (dot_count > norm_count * textord_noise_normratio)
        word_dud[word_index] = 1;
      else
        word_dud[word_index] = 0;
    } else {
      word_dud[word_index] = 0;
    }
    if (word_dud[word_index] == 2)
      dud_words++;
    else
      ok_words++;
    word_index++;
  }

  word_index = 0;
  for (word_it.mark_cycle_pt (); !word_it.cycled_list (); word_it.forward ()) {
    if (word_dud[word_index] == 2
    || (word_dud[word_index] == 1 && dud_words > ok_words)) {
//.........这里部分代码省略.........

bool Textord::clean_noise_from_row(          //remove empties
        ROW* row  //row to clean
) {
  bool testing_on;
  TBOX blob_box;                 //bounding box
  C_BLOB *blob;                  //current blob
  C_OUTLINE *outline;            //current outline
  WERD *word;                    //current word
  int32_t blob_size;             //biggest size
  int32_t trans_count = 0;       //no of transitions
  int32_t trans_threshold;       //noise tolerance
  int32_t dot_count;             //small objects
  int32_t norm_count;            //normal objects
  int32_t super_norm_count;      //real char-like
                                 //words of row
  WERD_IT word_it = row->word_list ();
  C_BLOB_IT blob_it;             //blob iterator
  C_OUTLINE_IT out_it;           //outline iterator

  testing_on = textord_test_y > row->base_line (textord_test_x)
               && textord_show_blobs
               && textord_test_y < row->base_line (textord_test_x) + row->x_height ();
  dot_count = 0;
  norm_count = 0;
  super_norm_count = 0;
  for (word_it.mark_cycle_pt (); !word_it.cycled_list (); word_it.forward ()) {
    word = word_it.data ();      //current word
                                 //blobs in word
    blob_it.set_to_list (word->cblob_list ());
    for (blob_it.mark_cycle_pt (); !blob_it.cycled_list ();
    blob_it.forward ()) {
      blob = blob_it.data ();
      if (!word->flag (W_DONT_CHOP)) {
                                 //get outlines
        out_it.set_to_list (blob->out_list ());
        for (out_it.mark_cycle_pt (); !out_it.cycled_list ();
        out_it.forward ()) {
          outline = out_it.data ();
          blob_box = outline->bounding_box ();
          blob_size =
            blob_box.width () >
            blob_box.height ()? blob_box.width () : blob_box.
            height();
          if (blob_size < textord_noise_sizelimit * row->x_height ())
            dot_count++;         //count smal outlines
          if (!outline->child ()->empty ()
            && blob_box.height () <
            (1 + textord_noise_syfract) * row->x_height ()
            && blob_box.height () >
            (1 - textord_noise_syfract) * row->x_height ()
            && blob_box.width () <
            (1 + textord_noise_sxfract) * row->x_height ()
            && blob_box.width () >
            (1 - textord_noise_sxfract) * row->x_height ())
            super_norm_count++;  //count smal outlines
        }
      }
      else
        super_norm_count++;
      blob_box = blob->bounding_box ();
      blob_size =
        blob_box.width () >
        blob_box.height ()? blob_box.width () : blob_box.height ();
      if (blob_size >= textord_noise_sizelimit * row->x_height ()
          && blob_size < row->x_height () * 2) {
        trans_threshold = blob_size / textord_noise_sizefraction;
        trans_count = blob->count_transitions (trans_threshold);
        if (trans_count < textord_noise_translimit)
          norm_count++;
      }
      else if (blob_box.height () > row->x_height () * 2
        && (!word_it.at_first () || !blob_it.at_first ()))
        dot_count += 2;
      if (testing_on) {
        tprintf
          ("Blob at (%d,%d) -> (%d,%d), ols=%d, tc=%d, bldiff=%g\n",
          blob_box.left (), blob_box.bottom (), blob_box.right (),
          blob_box.top (), blob->out_list ()->length (), trans_count,
          blob_box.bottom () - row->base_line (blob_box.left ()));
      }
    }
  }
  if (textord_noise_debug) {
    tprintf ("Row ending at (%d,%g):",
      blob_box.right (), row->base_line (blob_box.right ()));
    tprintf (" R=%g, dc=%d, nc=%d, %s\n",
      norm_count > 0 ? (float) dot_count / norm_count : 9999,
      dot_count, norm_count,
      dot_count > norm_count * textord_noise_normratio
      && dot_count > 2 ? "REJECTED" : "ACCEPTED");
  }
  return super_norm_count < textord_noise_sncount
    && dot_count > norm_count * textord_noise_rowratio && dot_count > 2;
}

    QString printWORD(WERD_RES *wordres,int height)
    {
        WERD *word = wordres->word;
        QString aux;
        BITS16 flags;                //flags about word
        flags.set_bit(W_SEGMENTED,word->flag(W_SEGMENTED));
        flags.set_bit(W_ITALIC,word->flag(W_ITALIC));
        flags.set_bit(W_BOL,word->flag(W_BOL));
        flags.set_bit(W_EOL,word->flag(W_EOL));
        flags.set_bit(W_NORMALIZED,word->flag(W_NORMALIZED));
        flags.set_bit(W_POLYGON,word->flag(W_POLYGON));
        flags.set_bit(W_LINEARC,word->flag(W_LINEARC));
        flags.set_bit(W_DONT_CHOP,word->flag(W_DONT_CHOP));
        flags.set_bit(W_REP_CHAR,word->flag(W_REP_CHAR));
        flags.set_bit(W_FUZZY_SP,word->flag(W_FUZZY_SP));
        flags.set_bit(W_FUZZY_NON,word->flag(W_FUZZY_NON));
        aux.append(QString("Blanks= %1\n").arg(word->space()));
        aux.append(printTBOX(word->bounding_box(),height,true));
        aux.append(QString("Correct= %1\n")
                    .arg(wordres->best_choice->unichar_string().string()));
        aux.append(QString("Flags = %1 = 0%2\n").arg(flags.val)
                  .arg(flags.val,0,8));
        aux.append (QString("   W_SEGMENTED = %1\n")
                  .arg(word->flag(W_SEGMENTED) ? "TRUE" : "FALSE "));
        aux.append (QString("   W_ITALIC = %1\n")
                  .arg(word->flag(W_ITALIC) ? "TRUE" : "FALSE "));
        aux.append (QString("   W_BOL = %1\n")
                  .arg(word->flag(W_BOL) ? "TRUE" : "FALSE "));
        aux.append (QString("   W_EOL = %1\n")
                  .arg(word->flag(W_EOL) ? "TRUE" : "FALSE "));
        aux.append (QString("   W_NORMALIZED = %1\n")
                  .arg(word->flag(W_NORMALIZED) ? "TRUE" : "FALSE "));
        aux.append (QString("   W_POLYGON = %1\n")
                  .arg(word->flag(W_POLYGON) ? "TRUE" : "FALSE "));
        aux.append (QString("   W_LINEARC = %1\n")
                  .arg(word->flag(W_LINEARC) ? "TRUE" : "FALSE "));
        aux.append (QString("   W_DONT_CHOP = %1\n")
                  .arg(word->flag(W_DONT_CHOP) ? "TRUE" : "FALSE "));
        aux.append (QString("   W_REP_CHAR = %1\n")
                  .arg(word->flag(W_REP_CHAR) ? "TRUE" : "FALSE "));
        aux.append (QString("   W_FUZZY_SP = %1\n")
                  .arg(word->flag(W_FUZZY_SP) ? "TRUE" : "FALSE "));
        aux.append (QString("   W_FUZZY_NON = %1\n")
                  .arg(word->flag(W_FUZZY_NON) ? "TRUE" : "FALSE "));
        aux.append(QString("Rejected cblob count = %1")
                    .arg(word->rej_cblob_list()->length()));
        return aux;

    }

void Textord::clean_noise_from_words(          //remove empties
                                     ROW *row  //row to clean
                                    ) {
  TBOX blob_box;                  //bounding box
  inT8 *word_dud;                //was it chucked
  C_BLOB *blob;                  //current blob
  C_OUTLINE *outline;            //current outline
  WERD *word;                    //current word
  inT32 blob_size;               //biggest size
  inT32 trans_count;             //no of transitions
  inT32 trans_threshold;         //noise tolerance
  inT32 dot_count;               //small objects
  inT32 norm_count;              //normal objects
  inT32 dud_words;               //number discarded
  inT32 ok_words;                //number remaining
  inT32 word_index;              //current word
                                 //words of row
  WERD_IT word_it = row->word_list ();
  C_BLOB_IT blob_it;             //blob iterator
  C_OUTLINE_IT out_it;           //outline iterator

  ok_words = word_it.length ();
  if (ok_words == 0 || textord_no_rejects)
    return;
  word_dud = (inT8 *) alloc_mem (ok_words * sizeof (inT8));
  dud_words = 0;
  ok_words = 0;
  word_index = 0;
  for (word_it.mark_cycle_pt (); !word_it.cycled_list (); word_it.forward ()) {
    word = word_it.data ();      //current word
    dot_count = 0;
    norm_count = 0;
                                 //blobs in word
    blob_it.set_to_list (word->cblob_list ());
    for (blob_it.mark_cycle_pt (); !blob_it.cycled_list ();
    blob_it.forward ()) {
      blob = blob_it.data ();
      if (!word->flag (W_DONT_CHOP)) {
                                 //get outlines
        out_it.set_to_list (blob->out_list ());
        for (out_it.mark_cycle_pt (); !out_it.cycled_list ();
        out_it.forward ()) {
          outline = out_it.data ();
          blob_box = outline->bounding_box ();
          blob_size =
            blob_box.width () >
            blob_box.height ()? blob_box.width () : blob_box.
            height();
          if (blob_size < textord_noise_sizelimit * row->x_height ())
            dot_count++;         //count smal outlines
          if (!outline->child ()->empty ()
            && blob_box.height () <
            (1 + textord_noise_syfract) * row->x_height ()
            && blob_box.height () >
            (1 - textord_noise_syfract) * row->x_height ()
            && blob_box.width () <
            (1 + textord_noise_sxfract) * row->x_height ()
            && blob_box.width () >
            (1 - textord_noise_sxfract) * row->x_height ())
            norm_count++;        //count smal outlines
        }
      }
      else
        norm_count++;
      blob_box = blob->bounding_box ();
      blob_size =
        blob_box.width () >
        blob_box.height ()? blob_box.width () : blob_box.height ();
      if (blob_size >= textord_noise_sizelimit * row->x_height ()
      && blob_size < row->x_height () * 2) {
        trans_threshold = blob_size / textord_noise_sizefraction;
        trans_count = blob->count_transitions (trans_threshold);
        if (trans_count < textord_noise_translimit)
          norm_count++;
      }
      else if (blob_box.height () > row->x_height () * 2
        && (!word_it.at_first () || !blob_it.at_first ()))
        dot_count += 2;
    }
    if (dot_count > 2) {
      if (dot_count > norm_count * textord_noise_normratio * 2)
        word_dud[word_index] = 2;
      else if (dot_count > norm_count * textord_noise_normratio)
        word_dud[word_index] = 1;
      else
        word_dud[word_index] = 0;
    }
    else
      word_dud[word_index] = 0;
    if (word_dud[word_index] == 2)
      dud_words++;
    else
      ok_words++;
    word_index++;
  }

  word_index = 0;
  for (word_it.mark_cycle_pt (); !word_it.cycled_list (); word_it.forward ()) {
    if (word_dud[word_index] == 2
    || (word_dud[word_index] == 1 && dud_words > ok_words)) {
//.........这里部分代码省略.........