Beispiel #1
0
// Fits splines to the textlines, or creates fake QSPLINES from the straight
// baselines that are already on the TO_ROWs.
// As a side-effect, computes the xheights of the rows and the block.
// Although x-height estimation is conceptually separate, it is part of
// detecting perspective distortion and therefore baseline fitting.
void BaselineBlock::FitBaselineSplines(bool enable_splines,
                                       bool show_final_rows,
                                       Textord* textord) {
  double gradient = tan(skew_angle_);
  FCOORD rotation(1.0f, 0.0f);

  if (enable_splines) {
    textord->make_spline_rows(block_, gradient, show_final_rows);
  } else {
    // Make a fake spline from the existing line.
    TBOX block_box= block_->block->bounding_box();
    TO_ROW_IT row_it = block_->get_rows();
    for (row_it.mark_cycle_pt(); !row_it.cycled_list(); row_it.forward()) {
      TO_ROW* row = row_it.data();
      inT32 xstarts[2] = { block_box.left(), block_box.right() };
      double coeffs[3] = { 0.0, row->line_m(), row->line_c() };
      row->baseline = QSPLINE(1, xstarts, coeffs);
      textord->compute_row_xheight(row, block_->block->classify_rotation(),
                                   row->line_m(), block_->line_size);
    }
  }
  textord->compute_block_xheight(block_, gradient);
  block_->block->set_xheight(block_->xheight);
  if (textord_restore_underlines)  // fix underlines
    restore_underlined_blobs(block_);
}
void tweak_row_baseline(ROW *row,
                        double blshift_maxshift,
                        double blshift_xfraction) {
  TBOX 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 < blshift_maxshift
        && blob_box.height () / row->x_height () > 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);
}