tordmain.cpp

一ＯＣＲ的相关资料。．希望对研究ＯＣＲ的朋友有所帮助．

      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;
      #ifndef SECURE_NAMES
      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 ()));
      }
      #endif
    }
  }
  #ifndef SECURE_NAMES
  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");
  }
  #endif
  return super_norm_count < textord_noise_sncount
    && dot_count > norm_count * textord_noise_rowratio && dot_count > 2;
}


/**********************************************************************
 * clean_noise_from_words
 *
 * Move blobs of words from rows of garbage into the reject blobs list.
 **********************************************************************/

void clean_noise_from_words(          //remove empties
                            ROW *row  //row to clean
                           ) {
  BOX 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)
    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) {
      word = word_it.data ();    //current word
                                 //rejected blobs
      blob_it.set_to_list (word->rej_cblob_list ());
                                 //move from blobs
      blob_it.add_list_after (word->cblob_list ());
    }
    word_index++;
  }
  free_mem(word_dud);
}


/**********************************************************************
 * tweak_row_baseline
 *
 * Shift baseline to fit the blobs more accurately where they are
 * close enough.
 **********************************************************************/

void tweak_row_baseline(          //remove empties
                        ROW *row  //row to clean
                       ) {
  BOX blob_box;                  //bounding box
  C_BLOB *blob;                  //current blob
  WERD *word;                    //current word
  INT32 blob_count;              //no of blobs
  INT32 src_index;               //source segment
  INT32 dest_index;              //destination segment
  INT32 *xstarts;                //spline segments
  double *coeffs;                //spline coeffs
  float ydiff;                   //baseline error
  float x_centre;                //centre of blob
                                 //words of row
  WERD_IT word_it = row->word_list ();
  C_BLOB_IT blob_it;             //blob iterator

  blob_count = 0;
  for (word_it.mark_cycle_pt (); !word_it.cycled_list (); word_it.forward ()) {
    word = word_it.data ();      //current word
                                 //get total blobs
    blob_count += word->cblob_list ()->length ();
  }
  if (blob_count == 0)
    return;
  xstarts =
    (INT32 *) alloc_mem ((blob_count + row->baseline.segments + 1) *
    sizeof (INT32));
  coeffs =
    (double *) alloc_mem ((blob_count + row->baseline.segments) * 3 *
    sizeof (double));

  src_index = 0;
  dest_index = 0;
  xstarts[0] = row->baseline.xcoords[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 ();
      blob_box = blob->bounding_box ();
      x_centre = (blob_box.left () + blob_box.right ()) / 2.0;
      ydiff = blob_box.bottom () - row->base_line (x_centre);
      if (ydiff < 0)
        ydiff = -ydiff / row->x_height ();
      else
        ydiff = ydiff / row->x_height ();
      if (ydiff < textord_blshift_maxshift
        && blob_box.height () / row->x_height () >
      textord_blshift_xfraction) {
        if (xstarts[dest_index] >= x_centre)
          xstarts[dest_index] = blob_box.left ();
        coeffs[dest_index * 3] = 0;
        coeffs[dest_index * 3 + 1] = 0;
        coeffs[dest_index * 3 + 2] = blob_box.bottom ();
        //shift it
        dest_index++;
        xstarts[dest_index] = blob_box.right () + 1;
      }
      else {
        if (xstarts[dest_index] <= x_centre) {
          while (row->baseline.xcoords[src_index + 1] <= x_centre
          && src_index < row->baseline.segments - 1) {
            if (row->baseline.xcoords[src_index + 1] >
            xstarts[dest_index]) {
              coeffs[dest_index * 3] =
                row->baseline.quadratics[src_index].a;
              coeffs[dest_index * 3 + 1] =
                row->baseline.quadratics[src_index].b;
              coeffs[dest_index * 3 + 2] =
                row->baseline.quadratics[src_index].c;
              dest_index++;
              xstarts[dest_index] =
                row->baseline.xcoords[src_index + 1];
            }
            src_index++;
          }
          coeffs[dest_index * 3] =
            row->baseline.quadratics[src_index].a;
          coeffs[dest_index * 3 + 1] =
            row->baseline.quadratics[src_index].b;
          coeffs[dest_index * 3 + 2] =
            row->baseline.quadratics[src_index].c;
          dest_index++;
          xstarts[dest_index] = row->baseline.xcoords[src_index + 1];
        }
      }
    }
  }
  while (src_index < row->baseline.segments
    && row->baseline.xcoords[src_index + 1] <= xstarts[dest_index])
    src_index++;
  while (src_index < row->baseline.segments) {
    coeffs[dest_index * 3] = row->baseline.quadratics[src_index].a;
    coeffs[dest_index * 3 + 1] = row->baseline.quadratics[src_index].b;
    coeffs[dest_index * 3 + 2] = row->baseline.quadratics[src_index].c;
    dest_index++;
    src_index++;
    xstarts[dest_index] = row->baseline.xcoords[src_index];
  }
                                 //turn to spline
  row->baseline = QSPLINE (dest_index, xstarts, coeffs);
  free_mem(xstarts);
  free_mem(coeffs);
}


/**********************************************************************
 * blob_y_order
 *
 * Sort function to sort blobs in y from page top.
 **********************************************************************/

INT32 blob_y_order(              //sort function
                   void *item1,  //items to compare
                   void *item2) {
                                 //converted ptr
  BLOBNBOX *blob1 = *(BLOBNBOX **) item1;
                                 //converted ptr
  BLOBNBOX *blob2 = *(BLOBNBOX **) item2;

  if (blob1->bounding_box ().bottom () > blob2->bounding_box ().bottom ())
    return -1;
  else if (blob1->bounding_box ().bottom () <
    blob2->bounding_box ().bottom ())
    return 1;
  else {
    if (blob1->bounding_box ().left () < blob2->bounding_box ().left ())
      return -1;
    else if (blob1->bounding_box ().left () >
      blob2->bounding_box ().left ())
      return 1;
    else
      return 0;
  }
}