topitch.cpp

一ＯＣＲ的相关资料。．希望对研究ＯＣＲ的朋友有所帮助．

}


/**********************************************************************
 * compute_pitch_sd
 *
 * Use a dp algorithm to fit the character cells and return the sd of
 * the cell size over the row.
 **********************************************************************/

float compute_pitch_sd(                            //find fp cells
                       TO_ROW *row,                //row to do
                       STATS *projection,          //vertical projection
                       INT16 projection_left,      //edge
                       INT16 projection_right,     //edge
                       float space_size,           //size of blank
                       float initial_pitch,        //guess at pitch
                       float &sp_sd,               //space sd
                       INT16 &mid_cuts,            //no of free cuts
                       ICOORDELT_LIST *row_cells,  //list of chop pts
                       BOOL8 testing_on,           //inidividual words
                       INT16 start,                //start of good range
                       INT16 end                   //end of good range
                      ) {
  INT16 occupation;              //no of cells in word.
                                 //blobs
  BLOBNBOX_IT blob_it = row->blob_list ();
  BLOBNBOX_IT start_it;          //start of word
  BLOBNBOX_IT plot_it;           //for plotting
  INT16 blob_count;              //no of blobs
  BOX blob_box;                  //bounding box
  BOX prev_box;                  //of super blob
  INT32 prev_right;              //of word sync
  int scale_factor;              //on scores for big words
  INT32 sp_count;                //spaces
  FPSEGPT_LIST seg_list;         //char cells
  FPSEGPT_IT seg_it;             //iterator
  INT16 segpos;                  //position of segment
  INT16 cellpos;                 //previous cell boundary
                                 //iterator
  ICOORDELT_IT cell_it = row_cells;
  ICOORDELT *cell;               //new cell
  double sqsum;                  //sum of squares
  double spsum;                  //of spaces
  double sp_var;                 //space error
  double word_sync;              //result for word
  INT32 total_count;             //total blobs

  if ((pitsync_linear_version & 3) > 1) {
    word_sync = compute_pitch_sd2 (row, projection, projection_left,
      projection_right, initial_pitch,
      occupation, mid_cuts, row_cells,
      testing_on, start, end);
    sp_sd = occupation;
    return word_sync;
  }
  mid_cuts = 0;
  cellpos = 0;
  total_count = 0;
  sqsum = 0;
  sp_count = 0;
  spsum = 0;
  prev_right = -1;
  if (blob_it.empty ())
    return space_size * 10;
#ifndef GRAPHICS_DISABLED
  if (testing_on && to_win > 0) {
    blob_box = blob_it.data ()->bounding_box ();
    projection->plot (to_win, projection_left,
      row->intercept (), 1.0f, -1.0f, CORAL);
  }
#endif
  start_it = blob_it;
  blob_count = 0;
  blob_box = box_next (&blob_it);//first blob
  blob_it.mark_cycle_pt ();
  do {
    for (; blob_count > 0; blob_count--)
      box_next(&start_it);
    do {
      prev_box = blob_box;
      blob_count++;
      blob_box = box_next (&blob_it);
    }
    while (!blob_it.cycled_list ()
      && blob_box.left () - prev_box.right () < space_size);
    plot_it = start_it;
    if (pitsync_linear_version & 3)
      word_sync =
        check_pitch_sync2 (&start_it, blob_count, (INT16) initial_pitch, 2,
        projection, projection_left, projection_right,
        row->xheight * textord_projection_scale,
        occupation, &seg_list, start, end);
    else
      word_sync =
        check_pitch_sync (&start_it, blob_count, (INT16) initial_pitch, 2,
        projection, &seg_list);
    if (testing_on) {
      tprintf ("Word ending at (%d,%d), len=%d, sync rating=%g, ",
        prev_box.right (), prev_box.top (),
        seg_list.length () - 1, word_sync);
      seg_it.set_to_list (&seg_list);
      for (seg_it.mark_cycle_pt (); !seg_it.cycled_list ();
      seg_it.forward ()) {
        if (seg_it.data ()->faked)
          tprintf ("(F)");
        tprintf ("%d, ", seg_it.data ()->position ());
        //                              tprintf("C=%g, s=%g, sq=%g\n",
        //                                      seg_it.data()->cost_function(),
        //                                      seg_it.data()->sum(),
        //                                      seg_it.data()->squares());
      }
      tprintf ("\n");
    }
#ifndef GRAPHICS_DISABLED
    if (textord_show_fixed_cuts && blob_count > 0 && to_win > 0)
      plot_fp_cells2(to_win, GOLDENROD, row, &seg_list);
#endif
    seg_it.set_to_list (&seg_list);
    if (prev_right >= 0) {
      sp_var = seg_it.data ()->position () - prev_right;
      sp_var -= floor (sp_var / initial_pitch + 0.5) * initial_pitch;
      sp_var *= sp_var;
      spsum += sp_var;
      sp_count++;
    }
    for (seg_it.mark_cycle_pt (); !seg_it.cycled_list (); seg_it.forward ()) {
      segpos = seg_it.data ()->position ();
      if (cell_it.empty () || segpos > cellpos + initial_pitch / 2) {
                                 //big gap
        while (!cell_it.empty () && segpos > cellpos + initial_pitch * 3 / 2) {
          cell = new ICOORDELT (cellpos + (INT16) initial_pitch, 0);
          cell_it.add_after_then_move (cell);
          cellpos += (INT16) initial_pitch;
        }
                                 //make new one
        cell = new ICOORDELT (segpos, 0);
        cell_it.add_after_then_move (cell);
        cellpos = segpos;
      }
      else if (segpos > cellpos - initial_pitch / 2) {
        cell = cell_it.data ();
                                 //average positions
        cell->set_x ((cellpos + segpos) / 2);
        cellpos = cell->x ();
      }
    }
    seg_it.move_to_last ();
    prev_right = seg_it.data ()->position ();
    if (textord_pitch_scalebigwords) {
      scale_factor = (seg_list.length () - 2) / 2;
      if (scale_factor < 1)
        scale_factor = 1;
    }
    else
      scale_factor = 1;
    sqsum += word_sync * scale_factor;
    total_count += (seg_list.length () - 1) * scale_factor;
    seg_list.clear ();
  }
  while (!blob_it.cycled_list ());
  sp_sd = sp_count > 0 ? sqrt (spsum / sp_count) : 0;
  return total_count > 0 ? sqrt (sqsum / total_count) : space_size * 10;
}


/**********************************************************************
 * compute_pitch_sd2
 *
 * Use a dp algorithm to fit the character cells and return the sd of
 * the cell size over the row.
 **********************************************************************/

float compute_pitch_sd2(                            //find fp cells
                        TO_ROW *row,                //row to do
                        STATS *projection,          //vertical projection
                        INT16 projection_left,      //edge
                        INT16 projection_right,     //edge
                        float initial_pitch,        //guess at pitch
                        INT16 &occupation,          //no of occupied cells
                        INT16 &mid_cuts,            //no of free cuts
                        ICOORDELT_LIST *row_cells,  //list of chop pts
                        BOOL8 testing_on,           //inidividual words
                        INT16 start,                //start of good range
                        INT16 end                   //end of good range
                       ) {
                                 //blobs
  BLOBNBOX_IT blob_it = row->blob_list ();
  BLOBNBOX_IT plot_it;
  INT16 blob_count;              //no of blobs
  BOX blob_box;                  //bounding box
  FPSEGPT_LIST seg_list;         //char cells
  FPSEGPT_IT seg_it;             //iterator
  INT16 segpos;                  //position of segment
                                 //iterator
  ICOORDELT_IT cell_it = row_cells;
  ICOORDELT *cell;               //new cell
  double word_sync;              //result for word

  mid_cuts = 0;
  if (blob_it.empty ()) {
    occupation = 0;
    return initial_pitch * 10;
  }
#ifndef GRAPHICS_DISABLED
  if (testing_on && to_win > 0) {
    projection->plot (to_win, projection_left,
      row->intercept (), 1.0f, -1.0f, CORAL);
  }
#endif
  blob_count = 0;
  blob_it.mark_cycle_pt ();
  do {
                                 //first blob
    blob_box = box_next (&blob_it);
    blob_count++;
  }
  while (!blob_it.cycled_list ());
  plot_it = blob_it;
  word_sync = check_pitch_sync2 (&blob_it, blob_count, (INT16) initial_pitch,
    2, projection, projection_left,
    projection_right,
    row->xheight * textord_projection_scale,
    occupation, &seg_list, start, end);
  if (testing_on) {
    tprintf ("Row ending at (%d,%d), len=%d, sync rating=%g, ",
      blob_box.right (), blob_box.top (),
      seg_list.length () - 1, word_sync);
    seg_it.set_to_list (&seg_list);
    for (seg_it.mark_cycle_pt (); !seg_it.cycled_list (); seg_it.forward ()) {
      if (seg_it.data ()->faked)
        tprintf ("(F)");
      tprintf ("%d, ", seg_it.data ()->position ());
      //                              tprintf("C=%g, s=%g, sq=%g\n",
      //                                      seg_it.data()->cost_function(),
      //                                      seg_it.data()->sum(),
      //                                      seg_it.data()->squares());
    }
    tprintf ("\n");
  }
#ifndef GRAPHICS_DISABLED
  if (textord_show_fixed_cuts && blob_count > 0 && to_win > 0)
    plot_fp_cells2(to_win, GOLDENROD, row, &seg_list);
#endif
  seg_it.set_to_list (&seg_list);
  for (seg_it.mark_cycle_pt (); !seg_it.cycled_list (); seg_it.forward ()) {
    segpos = seg_it.data ()->position ();
                                 //make new one
    cell = new ICOORDELT (segpos, 0);
    cell_it.add_after_then_move (cell);
    if (seg_it.at_last ())
      mid_cuts = seg_it.data ()->cheap_cuts ();
  }
  seg_list.clear ();
  return occupation > 0 ? sqrt (word_sync / occupation) : initial_pitch * 10;
}


/**********************************************************************
 * print_pitch_sd
 *
 * Use a dp algorithm to fit the character cells and return the sd of
 * the cell size over the row.
 **********************************************************************/

void print_pitch_sd(                        //find fp cells
                    TO_ROW *row,            //row to do
                    STATS *projection,      //vertical projection
                    INT16 projection_left,  //edges //size of blank
                    INT16 projection_right,
                    float space_size,
                    float initial_pitch     //guess at pitch
                   ) {
  const char *res2;              //pitch result
  INT16 occupation;              //used cells
  float sp_sd;                   //space sd
                                 //blobs
  BLOBNBOX_IT blob_it = row->blob_list ();
  BLOBNBOX_IT start_it;          //start of word
  BLOBNBOX_IT row_start;         //start of row
  INT16 blob_count;              //no of blobs
  INT16 total_blob_count;        //total blobs in line
  BOX blob_box;                  //bounding box
  BOX prev_box;                  //of super blob
  INT32 prev_right;              //of word sync
  int scale_factor;              //on scores for big words
  INT32 sp_count;                //spaces
  FPSEGPT_LIST seg_list;         //char cells
  FPSEGPT_IT seg_it;             //iterator
  double sqsum;                  //sum of squares
  double spsum;                  //of spaces
  double sp_var;                 //space error
  double word_sync;              //result for word
  double total_count;            //total cuts

  if (blob_it.empty ())
    return;
  row_start = blob_it;
  total_blob_count = 0;

  total_count = 0;
  sqsum = 0;
  sp_count = 0;
  spsum = 0;
  prev_right = -1;
  blob_it = row_start;
  start_it = blob_it;
  blob_count = 0;
  blob_box = box_next (&blob_it);//first blob
  blob_it.mark_cycle_pt ();
  do {
    for (; blob_count > 0; blob_count--)
      box_next(&start_it);
    do {
      prev_box = blob_box;
      blob_count++;
      blob_box = box_next (&blob_it);
    }
    while (!blob_it.cycled_list ()
      && blob_box.left () - prev_box.right () < space_size);
    word_sync =
      check_pitch_sync2 (&start_it, blob_count, (INT16) initial_pitch, 2,
      projection, projection_left, projection_right,
      row->xheight * textord_projection_scale,
      occupation, &seg_list, 0, 0);
    total_blob_count += blob_count;
    seg_it.set_to_list (&seg_list);
    if (prev_right >= 0) {
      sp_var = seg_it.data ()->position () - prev_right;
      sp_var -= floor (sp_var / initial_pitch + 0.5) * initial_pitch;
      sp_var *= sp_var;
      spsum += sp_var;
      sp_count++;
    }
    seg_it.move_to_last ();
    prev_right = seg_it.data ()->position ();
    if (textord_pitch_scalebigwords) {
      scale_factor = (seg_list.length () - 2) / 2;
      if (scale_factor < 1)
        scale_factor = 1;
    }
    else
      scale_factor = 1;
    sqsum += word_sync * scale_factor;
    total_count += (seg_list.length () - 1) * scale_factor;
    seg_list.clear ();
  }
  while (!blob_it.cycled_list ());
  sp_sd = sp