werd.cpp

一ＯＣＲ的相关资料。．希望对研究ＯＣＲ的朋友有所帮助．

    for (cblob_it.mark_cycle_pt ();
    !cblob_it.cycled_list (); cblob_it.forward ())
  cblob_it.data ()->move (vec);
}


/**********************************************************************
 * WERD::scale
 *
 * Scale WERD by multiplier
 **********************************************************************/

void WERD::scale(               // scale WERD
                 const float f  // by multiplier
                ) {
  PBLOB_IT blob_it ((PBLOB_LIST *) & cblobs);
  // blob iterator
  //      LARC_BLOB_IT                            lblob_it((LARC_BLOB_LIST*)&cblobs);

  if (flags.bit (W_POLYGON))
    for (blob_it.mark_cycle_pt ();
    !blob_it.cycled_list (); blob_it.forward ())
  blob_it.data ()->scale (f);
  //      else if (flags.bit(W_LINEARC))
  //              for (lblob_it.mark_cycle_pt();
  //                              !lblob_it.cycled_list();
  //                              lblob_it.forward() )
  //                      lblob_it.data()->scale( f );
  else
    CANT_SCALE_EDGESTEPS.error ("WERD::scale", ABORT, NULL);
}


/**********************************************************************
 * WERD::join_on
 *
 * Join other word onto this one. Delete the old word.
 **********************************************************************/

void WERD::join_on(              // join WERD
                   WERD *&other  //other word
                  ) {
  PBLOB_IT blob_it ((PBLOB_LIST *) & cblobs);
  // blob iterator
  PBLOB_IT src_it ((PBLOB_LIST *) & other->cblobs);
  C_BLOB_IT rej_cblob_it(&rej_cblobs); 
  C_BLOB_IT src_rej_it (&other->rej_cblobs);

  while (!src_it.empty ()) {
    blob_it.add_to_end (src_it.extract ());
    src_it.forward ();
  }
  while (!src_rej_it.empty ()) {
    rej_cblob_it.add_to_end (src_rej_it.extract ());
    src_rej_it.forward ();
  }
}


/**********************************************************************
 * WERD::copy_on
 *
 * Copy blobs from other word onto this one.
 **********************************************************************/

void WERD::copy_on(              //copy blobs
                   WERD *&other  //from other
                  ) {
  if (flags.bit (W_POLYGON)) {
    PBLOB_IT blob_it ((PBLOB_LIST *) & cblobs);
    // blob iterator
    PBLOB_LIST blobs;

    blobs.deep_copy ((PBLOB_LIST *) (&other->cblobs));
    blob_it.move_to_last ();
    blob_it.add_list_after (&blobs);
  }
  //      else if (flags.bit(W_LINEARC))
  //      {
  //              LARC_BLOB_IT                            larc_blob_it( (LARC_BLOB_LIST*)&cblobs );
  //              LARC_BLOB_LIST                          larc_blobs;

  //              larc_blobs.deep_copy((LARC_BLOB_LIST*)(&other->cblobs));
  //    larc_blob_it.move_to_last();
  //              larc_blob_it.add_list_after( &larc_blobs );
  //      }
  else {
    C_BLOB_IT c_blob_it(&cblobs); 
    C_BLOB_LIST c_blobs;

    c_blobs.deep_copy (&other->cblobs);
    c_blob_it.move_to_last ();
    c_blob_it.add_list_after (&c_blobs);
  }
  if (!other->rej_cblobs.empty ()) {
    C_BLOB_IT rej_c_blob_it(&rej_cblobs); 
    C_BLOB_LIST new_rej_c_blobs;

    new_rej_c_blobs.deep_copy (&other->rej_cblobs);
    rej_c_blob_it.move_to_last ();
    rej_c_blob_it.add_list_after (&new_rej_c_blobs);
  }
}


/**********************************************************************
 * WERD::baseline_normalise
 *
 * Baseline Normalise the word in Tesseract style.  (I.e origin at centre of
 * word at bottom. x-height region scaled to region y =
 * (bln_baseline_offset)..(bln_baseline_offset + bln_x_height)
 * - usually 64..192)
 **********************************************************************/

void WERD::baseline_normalise(                // Tess style BL Norm
                              ROW *row,
                              DENORM *denorm  //antidote
                             ) {
  baseline_normalise_x (row, row->x_height (), denorm);
  //Use standard x ht
}


/**********************************************************************
 * WERD::baseline_normalise_x
 *
 * Baseline Normalise the word in Tesseract style.  (I.e origin at centre of
 * word at bottom. x-height region scaled to region y =
 * (bln_baseline_offset)..(bln_baseline_offset + bln_x_height)
 * - usually 64..192)
 *  USE A SPECIFIED X-HEIGHT - NOT NECESSARILY THE ONE IN row
 **********************************************************************/

void WERD::baseline_normalise_x(                 // Tess style BL Norm
                                ROW *row,
                                float x_height,  //non standard value
                                DENORM *denorm   //antidote
                               ) {
  BOOL8 using_row;               //as baseline
  float blob_x_centre;           //middle of blob
  float blob_offset;             //bottom miss
  float top_offset;              //top miss
  float blob_x_height;           //xh for this blob
  INT16 segments;                //no of segments
  INT16 segment;                 //current segment
  DENORM_SEG *segs;              //array of segments
  float mean_x;                  //mean xheight
  INT32 x_count;                 //no of xs
  BOX word_box = bounding_box ();//word bounding box
  BOX blob_box;                  //blob bounding box
  PBLOB_IT blob_it ((PBLOB_LIST *) & cblobs);
  // blob iterator
  PBLOB *blob;
  LLSQ line;                     //fitted line
  double line_m, line_c;         //fitted line
                                 //inverse norm
  DENORM antidote (word_box.left () +

    (word_box.right () - word_box.left ()) / 2.0,
    bln_x_height / x_height, row);

  if (!flags.bit (W_POLYGON)) {
    WRONG_WORD.error ("WERD::baseline_normalise", ABORT,
      "Need to poly approx");
  }

  if (flags.bit (W_NORMALIZED)) {
    WRONG_WORD.error ("WERD::baseline_normalise", ABORT,
      "Baseline unnormalised");
  }

  if (bln_numericmode) {
    segs = new DENORM_SEG[blob_it.length ()];
    segments = 0;
    float factor;  // For scaling to baseline normalised size.
    for (blob_it.mark_cycle_pt (); !blob_it.cycled_list ();
    blob_it.forward ()) {
      blob = blob_it.data ();
      blob_box = blob->bounding_box ();
      blob->move (FCOORD (-antidote.origin (),
        -blob_box.bottom ()));
      factor = bln_x_height * 4.0f / (3 * blob_box.height ());
      // Constrain the scale factor as target numbers should be either
      // cap height already or xheight.
      if (factor < antidote.scale())
        factor = antidote.scale();
      else if (factor > antidote.scale() * 1.5f)
        factor = antidote.scale() * 1.5f;
      blob->scale (factor);
      blob->move (FCOORD (0.0, bln_baseline_offset));
      segs[segments].xstart = blob->bounding_box().left();
      segs[segments].ycoord = blob_box.bottom();
      segs[segments++].scale_factor = factor;
    }
    antidote = DENORM (antidote.origin (), antidote.scale (),
      0.0f, 0.0f, segments, segs, true, row);
    delete [] segs;

                                 //Repeat for rej blobs
    blob_it.set_to_list ((PBLOB_LIST *) & rej_cblobs);
    for (blob_it.mark_cycle_pt (); !blob_it.cycled_list ();
    blob_it.forward ()) {
      blob = blob_it.data ();
      blob_box = blob->bounding_box ();
      blob->move (FCOORD (-antidote.origin (),
                          -blob_box.bottom ()));
      blob->scale (bln_x_height * 4.0f / (3 * blob_box.height ()));
      blob->move (FCOORD (0.0, bln_baseline_offset));
    }
  }
  else if (bln_blshift_maxshift < 0) {
    for (blob_it.mark_cycle_pt (); !blob_it.cycled_list ();
    blob_it.forward ()) {
      blob = blob_it.data ();
      blob_box = blob->bounding_box ();
      blob_x_centre = blob_box.left () +
        (blob_box.right () - blob_box.left ()) / 2.0;
      blob->move (FCOORD (-antidote.origin (),
        -(row->base_line (blob_x_centre))));
      blob->scale (antidote.scale ());
      blob->move (FCOORD (0.0, bln_baseline_offset));
    }

                                 //Repeat for rej blobs
    blob_it.set_to_list ((PBLOB_LIST *) & rej_cblobs);
    for (blob_it.mark_cycle_pt (); !blob_it.cycled_list ();
    blob_it.forward ()) {
      blob = blob_it.data ();
      blob_box = blob->bounding_box ();
      blob_x_centre = blob_box.left () +
        (blob_box.right () - blob_box.left ()) / 2.0;
      blob->move (FCOORD (-antidote.origin (),
        -(row->base_line (blob_x_centre))));
      blob->scale (antidote.scale ());
      blob->move (FCOORD (0.0, bln_baseline_offset));
    }

  }
  else {
    mean_x = x_height;
    x_count = 1;
    segs = new DENORM_SEG[blob_it.length ()];
    segments = 0;
    for (blob_it.mark_cycle_pt (); !blob_it.cycled_list ();
    blob_it.forward ()) {
      blob = blob_it.data ();
      blob_box = blob->bounding_box ();
      if (blob_box.height () > bln_blshift_xfraction * x_height) {
        blob_x_centre = blob_box.left () +
          (blob_box.right () - blob_box.left ()) / 2.0;
        blob_offset =
          blob_box.bottom () - row->base_line (blob_x_centre);
        top_offset = blob_offset + blob_box.height () - x_height - 1;
        blob_x_height = top_offset + x_height;
        if (top_offset < 0)
          top_offset = -top_offset;
        if (blob_offset < 0)
          blob_offset = -blob_offset;
        if (blob_offset < bln_blshift_maxshift * x_height) {
          segs[segments].ycoord = blob_box.bottom ();
          line.add (blob_x_centre, blob_box.bottom ());
          if (top_offset < bln_blshift_maxshift * x_height) {
            segs[segments].scale_factor = blob_box.height () - 1.0f;
            x_count++;
          }
          else
            segs[segments].scale_factor = 0.0f;
          //fix it later
        }
        else {
                                 //not a goer
          segs[segments].ycoord = -MAX_INT32;
          if (top_offset < bln_blshift_maxshift * x_height) {
            segs[segments].scale_factor = blob_x_height;
            x_count++;
          }
          else
            segs[segments].scale_factor = 0.0f;
          //fix it later
        }
      }
      else {
        segs[segments].scale_factor = 0.0f;
        segs[segments].ycoord = -MAX_INT32;
      }
      segs[segments].xstart = blob_box.left ();
      segments++;
    }
    using_row = line.count () <= 1;
    if (!using_row) {
      line_m = line.m ();
      line_c = line.c (line_m);
    }
    else
      line_m = line_c = 0;
    segments = 0;
    for (blob_it.mark_cycle_pt (); !blob_it.cycled_list ();
    blob_it.forward ()) {
      blob = blob_it.data ();
      blob_box = blob->bounding_box ();
      blob_x_centre = blob_box.left () +
        (blob_box.right () - blob_box.left ()) / 2.0;
      if (segs[segments].ycoord == -MAX_INT32
      && segs[segments].scale_factor != 0 && !using_row) {
        blob_offset = line_m * blob_x_centre + line_c;
        segs[segments].scale_factor = blob_box.top () - blob_offset;
      }
      if (segs[segments].scale_factor != 0)
        mean_x += segs[segments].scale_factor;
      segments++;
    }
    mean_x /= x_count;
    //              printf("mean x=%g, count=%d, line_m=%g, line_c=%g\n",
    //                      mean_x,x_count,line_m,line_c);
    segments = 0;
    for (blob_it.mark_cycle_pt (); !blob_it.cycled_list ();
    blob_it.forward ()) {
      blob = blob_it.data ();
      blob_box = blob->bounding_box ();
      blob_x_centre = blob_box.left () +
        (blob_box.right () - blob_box.left ()) / 2.0;
      if (segs[segments].ycoord != -MAX_INT32)
        blob_offset = (float) segs[segments].ycoord;
      else if (using_row)
        blob_offset = row->base_line (blob_x_centre);
      else
        blob_offset = line_m * blob_x_centre + line_c;
      if (segs[segments].scale_factor == 0)
        segs[segments].scale_factor = mean_x;
      segs[segments].scale_factor =
        bln_x_height / segs[segments].scale_factor;
      //                      printf("Blob sf=%g, top=%d, bot=%d, base=%g\n",
      //                              segs[segments].scale_factor,blob_box.top(),