charsample.cpp

来自「一ＯＣＲ的相关资料。．希望对研究ＯＣＲ的朋友有所帮助．」· C++ 代码 · 共 699 行 · 第 1/2 页

  ysize = 0;
  ch = '\0';
  nsamples = 0;
  proto_data = NULL;
  proto = NULL;
}


CHAR_PROTO::CHAR_PROTO(INT32 x_size,
                       INT32 y_size,
                       INT32 n_samples,
                       float initial_value,
                       char c) {
  INT32 x;
  INT32 y;

  xsize = x_size;
  ysize = y_size;
  ch = c;
  nsamples = n_samples;

  ALLOC_2D_ARRAY(xsize, ysize, proto_data, proto, float); 

  for (y = 0; y < ysize; y++)
    for (x = 0; x < xsize; x++)
      proto[x][y] = initial_value;
}


CHAR_PROTO::CHAR_PROTO(CHAR_SAMPLE *sample) { 
  INT32 x;
  INT32 y;
  IMAGELINE imline_s;

  if (sample->image () == NULL) {
    xsize = 0;
    ysize = 0;
    ch = '\0';
    nsamples = 0;
    proto_data = NULL;
    proto = NULL;
  }
  else {
    ch = sample->character ();
    xsize = sample->image ()->get_xsize ();
    ysize = sample->image ()->get_ysize ();
    nsamples = 1;

    ALLOC_2D_ARRAY(xsize, ysize, proto_data, proto, float); 

    for (y = 0; y < ysize; y++) {
      sample->image ()->fast_get_line (0, y, xsize, &imline_s);
      for (x = 0; x < xsize; x++)
        if (imline_s.pixels[x] == BINIM_WHITE)
          proto[x][y] = 1.0;
      else
        proto[x][y] = -1.0;
    }
  }
}


CHAR_PROTO::~CHAR_PROTO () {
  if (proto_data != NULL)
    FREE_2D_ARRAY(proto_data, proto); 
}


float CHAR_PROTO::match_sample(CHAR_SAMPLE *test_sample) { 
  CHAR_PROTO *test_proto;
  float score;

  if (test_sample->image () != NULL) {
    test_proto = new CHAR_PROTO (test_sample);
    if (xsize > test_proto->x_size ())
      score = this->match (test_proto);
    else {
      demo_word = -demo_word;    // Flag different call
      score = test_proto->match (this);
    }
  }
  else
    return BAD_SCORE;

  delete test_proto;

  return score;
}


float CHAR_PROTO::match(CHAR_PROTO *test_proto) { 
  INT32 xsize2 = test_proto->x_size ();
  INT32 y_size;
  INT32 y_size2;
  INT32 x_offset;
  INT32 y_offset;
  INT32 x;
  INT32 y;
  CHAR_PROTO *match_proto;
  float score;
  float sum = 0.0;

  ASSERT_HOST (xsize >= xsize2);

  x_offset = (xsize - xsize2) / 2;

  if (ysize < test_proto->y_size ()) {
    y_size = test_proto->y_size ();
    y_size2 = ysize;
    y_offset = (y_size - y_size2) / 2;

    match_proto = new CHAR_PROTO (xsize,
      y_size,
      nsamples * test_proto->n_samples (),
      0, '\0');

    for (y = 0; y < y_offset; y++) {
      for (x = 0; x < xsize2; x++) {
        match_proto->data ()[x + x_offset][y] =
          test_proto->data ()[x][y] * nsamples;
        sum += match_proto->data ()[x + x_offset][y];
      }
    }

    for (y = y_offset + y_size2; y < y_size; y++) {
      for (x = 0; x < xsize2; x++) {
        match_proto->data ()[x + x_offset][y] =
          test_proto->data ()[x][y] * nsamples;
        sum += match_proto->data ()[x + x_offset][y];
      }
    }

    for (y = y_offset; y < y_offset + y_size2; y++) {
      for (x = 0; x < x_offset; x++) {
        match_proto->data ()[x][y] = proto[x][y - y_offset] *
          test_proto->n_samples ();
        sum += match_proto->data ()[x][y];
      }

      for (x = x_offset + xsize2; x < xsize; x++) {
        match_proto->data ()[x][y] = proto[x][y - y_offset] *
          test_proto->n_samples ();
        sum += match_proto->data ()[x][y];
      }

      for (x = x_offset; x < x_offset + xsize2; x++) {
        match_proto->data ()[x][y] =
          proto[x][y - y_offset] * test_proto->data ()[x - x_offset][y];
        sum += match_proto->data ()[x][y];
      }
    }
  }
  else {
    y_size = ysize;
    y_size2 = test_proto->y_size ();
    y_offset = (y_size - y_size2) / 2;

    match_proto = new CHAR_PROTO (xsize,
      y_size,
      nsamples * test_proto->n_samples (),
      0, '\0');

    for (y = 0; y < y_offset; y++)
    for (x = 0; x < xsize; x++) {
      match_proto->data ()[x][y] =
        proto[x][y] * test_proto->n_samples ();
      sum += match_proto->data ()[x][y];
    }

    for (y = y_offset + y_size2; y < y_size; y++)
    for (x = 0; x < xsize; x++) {
      match_proto->data ()[x][y] =
        proto[x][y] * test_proto->n_samples ();
      sum += match_proto->data ()[x][y];
    }

    for (y = y_offset; y < y_offset + y_size2; y++) {
      for (x = 0; x < x_offset; x++) {
        match_proto->data ()[x][y] =
          proto[x][y] * test_proto->n_samples ();
        sum += match_proto->data ()[x][y];
      }

      for (x = x_offset + xsize2; x < xsize; x++) {
        match_proto->data ()[x][y] =
          proto[x][y] * test_proto->n_samples ();
        sum += match_proto->data ()[x][y];
      }

      for (x = x_offset; x < x_offset + xsize2; x++) {
        match_proto->data ()[x][y] = proto[x][y] *
          test_proto->data ()[x - x_offset][y - y_offset];
        sum += match_proto->data ()[x][y];
      }
    }
  }

  score = (1.0 - sum /
    (xsize * y_size * nsamples * test_proto->n_samples ()));

  if (tessedit_mm_debug) {
    if (score < 0) {
      tprintf ("Match score %f\n", score);
      tprintf ("x: %d, y: %d, ns: %d, nt: %d, dx %d, dy: %d\n",
        xsize, y_size, nsamples, test_proto->n_samples (),
        x_offset, y_offset);
      for (y = 0; y < y_size; y++) {
        tprintf ("\n%d", y);
        for (x = 0; x < xsize; x++)
          tprintf ("\t%d", match_proto->data ()[x][y]);

      }
      tprintf ("\n");
      fflush(debug_fp); 
    }
  }

#ifndef GRAPHICS_DISABLED
  if (tessedit_display_mm) {
    tprintf ("Match score %f\n", score);
    display_images (this->make_image (),
      test_proto->make_image (), match_proto->make_image ());
  }
  else if (demo_word != 0) {
    if (demo_word > 0)
      display_image (test_proto->make_image (), "Test sample",
        300, 400, FALSE);
    else
      display_image (this->make_image (), "Test sample", 300, 400, FALSE);

    display_image (match_proto->make_image (), "Best match",
      700, 400, TRUE);
  }
#endif

  delete match_proto;

  return score;
}


void CHAR_PROTO::enlarge_prototype(INT32 new_xsize, INT32 new_ysize) { 
  float *old_proto_data = proto_data;
  float **old_proto = proto;
  INT32 old_xsize = xsize;
  INT32 old_ysize = ysize;
  INT32 x_offset;
  INT32 y_offset;
  INT32 x;
  INT32 y;

  ASSERT_HOST (new_xsize >= xsize && new_ysize >= ysize);

  xsize = new_xsize;
  ysize = new_ysize;
  ALLOC_2D_ARRAY(xsize, ysize, proto_data, proto, float); 
  x_offset = (xsize - old_xsize) / 2;
  y_offset = (ysize - old_ysize) / 2;

  for (y = 0; y < y_offset; y++)
    for (x = 0; x < xsize; x++)
      proto[x][y] = nsamples;

  for (y = y_offset + old_ysize; y < ysize; y++)
    for (x = 0; x < xsize; x++)
      proto[x][y] = nsamples;

  for (y = y_offset; y < y_offset + old_ysize; y++) {
    for (x = 0; x < x_offset; x++)
      proto[x][y] = nsamples;

    for (x = x_offset + old_xsize; x < xsize; x++)
      proto[x][y] = nsamples;

    for (x = x_offset; x < x_offset + old_xsize; x++)
      proto[x][y] = old_proto[x - x_offset][y - y_offset];
  }

  FREE_2D_ARRAY(old_proto_data, old_proto); 
}


void CHAR_PROTO::add_sample(CHAR_SAMPLE *sample) { 
  INT32 x_offset;
  INT32 y_offset;
  INT32 x;
  INT32 y;
  IMAGELINE imline_s;
  INT32 sample_xsize = sample->image ()->get_xsize ();
  INT32 sample_ysize = sample->image ()->get_ysize ();

  x_offset = (xsize - sample_xsize) / 2;
  y_offset = (ysize - sample_ysize) / 2;

  ASSERT_HOST (x_offset >= 0 && y_offset >= 0);

  for (y = 0; y < y_offset; y++)
    for (x = 0; x < xsize; x++)
      proto[x][y]++;             // Treat pixels outside the
  // range as white
  for (y = y_offset + sample_ysize; y < ysize; y++)
    for (x = 0; x < xsize; x++)
      proto[x][y]++;

  for (y = y_offset; y < y_offset + sample_ysize; y++) {
    sample->image ()->fast_get_line (0,
      y - y_offset, sample_xsize, &imline_s);
    for (x = x_offset; x < x_offset + sample_xsize; x++) {
      if (imline_s.pixels[x - x_offset] == BINIM_WHITE)
        proto[x][y]++;
      else
        proto[x][y]--;
    }

    for (x = 0; x < x_offset; x++)
      proto[x][y]++;

    for (x = x_offset + sample_xsize; x < xsize; x++)
      proto[x][y]++;
  }

  nsamples++;
}


IMAGE *CHAR_PROTO::make_image() { 
  IMAGE *image;
  IMAGELINE imline_p;
  INT32 x;
  INT32 y;

  ASSERT_HOST (nsamples != 0);

  image = new (IMAGE);
  image->create (xsize, ysize, 8);

  for (y = 0; y < ysize; y++) {
    image->fast_get_line (0, y, xsize, &imline_p);

    for (x = 0; x < xsize; x++) {
      imline_p.pixels[x] = 128 +
        (UINT8) ((proto[x][y] * 128.0) / (0.00001 + nsamples));
    }

    image->fast_put_line (0, y, xsize, &imline_p);
  }
  return image;
}