Method Crop

deps/tesseract/cube/char_samp.cpp:338–368

Crop the char samp such that there are no white spaces on any side. The norm_top_ and norm_bottom_ fields are the character top/bottom with respect to whatever context the character is being recognized in (e.g. word bounding box) normalized to a standard size of 255. Here they default to 0 and 255 (word box boundaries), but since they are context dependent, they may need to be reset by the calling

// since they are context dependent, they may need to be reset by the
// calling function.
CharSamp *CharSamp::Crop() {
  // get the dimesions of the cropped img
  int cropped_left = 0;
  int cropped_top = 0;
  int cropped_wid = wid_;
  int cropped_hgt = hgt_;
  Bmp8::Crop(&cropped_left, &cropped_top,
             &cropped_wid, &cropped_hgt);

  if (cropped_wid == 0 || cropped_hgt == 0) {
    return NULL;
  }
  // create the cropped char samp
  CharSamp *cropped_samp = new CharSamp(left_ + cropped_left,
                                        top_ + cropped_top,
                                        cropped_wid, cropped_hgt);
  cropped_samp->SetLabel(label32_);
  cropped_samp->SetFirstChar(first_char_);
  cropped_samp->SetLastChar(last_char_);
  // the following 3 fields may/should be reset by the calling function
  // using context information, i.e., location of character box
  // w.r.t. the word bounding box
  cropped_samp->SetNormAspectRatio(255 *
                                   cropped_wid / (cropped_wid + cropped_hgt));
  cropped_samp->SetNormTop(0);
  cropped_samp->SetNormBottom(255);

  // copy the bitmap to the cropped img
  Copy(cropped_left, cropped_top, cropped_wid, cropped_hgt, cropped_samp);
  return cropped_samp;
}

// segment the char samp to connected components
// based on contiguity and vertical pixel density histogram