tesseract::ParagraphModelSmearer Class Reference

#include <paragraphs_internal.h>

Public Member Functions
	ParagraphModelSmearer (GenericVector< RowScratchRegisters > *rows, int row_start, int row_end, ParagraphTheory *theory)
void	Smear ()

Detailed Description

Definition at line 236 of file paragraphs_internal.h.

Constructor & Destructor Documentation

ParagraphModelSmearer()

tesseract::ParagraphModelSmearer::ParagraphModelSmearer	(	GenericVector< RowScratchRegisters > *	rows,
		int	row_start,
		int	row_end,
		ParagraphTheory *	theory
	)

Definition at line 1335 of file paragraphs.cpp.

                                                                          {
  SetOfModels no_models;
  if (row_start < row_start_) row_start = row_start_;
  if (row_end > row_end_) row_end = row_end_;
 
  for (int row = (row_start > 0) ? row_start - 1 : row_start; row < row_end;
       row++) {
    if ((*rows_)[row].ri_->num_words == 0) {
      OpenModels(row + 1) = no_models;
    } else {
      SetOfModels &opened = OpenModels(row);
      (*rows_)[row].StartHypotheses(&opened);
 

Member Function Documentation

Smear()

void tesseract::ParagraphModelSmearer::Smear

(

)

Definition at line 1382 of file paragraphs.cpp.

                                                   {
      switch (OpenModels(i)[m]->justification()) {
        case JUSTIFICATION_LEFT: left_align_open = true; break;
        case JUSTIFICATION_RIGHT: right_align_open = true; break;
        default: left_align_open = right_align_open = true;
      }
    }
    // Step Two:
    //   Use that knowledge to figure out if this row is likely to
    //   start a paragraph.
    bool likely_start;
    if (i == 0) {
      likely_start = true;
    } else {
      if ((left_align_open && right_align_open) ||
          (!left_align_open && !right_align_open)) {
        likely_start = LikelyParagraphStart((*rows_)[i - 1], row,
                                            JUSTIFICATION_LEFT) ||
                       LikelyParagraphStart((*rows_)[i - 1], row,
                                            JUSTIFICATION_RIGHT);
      } else if (left_align_open) {
        likely_start = LikelyParagraphStart((*rows_)[i - 1], row,
                                            JUSTIFICATION_LEFT);
      } else {
        likely_start = LikelyParagraphStart((*rows_)[i - 1], row,
                                            JUSTIFICATION_RIGHT);
      }
    }
 
    // Step Three:
    //   If this text line seems like an obvious first line of an
    //   open model, or an obvious continuation of an existing
    //   modelled paragraph, mark it up.
    if (likely_start) {
      // Add Start Hypotheses for all Open models that fit.
      for (int m = 0; m < OpenModels(i).size(); m++) {
        if (ValidFirstLine(rows_, i, OpenModels(i)[m])) {
          row.AddStartLine(OpenModels(i)[m]);
        }
      }
    } else {
      // Add relevant body line hypotheses.
      SetOfModels last_line_models;
      if (i > 0) {
        (*rows_)[i - 1].StrongHypotheses(&last_line_models);
      } else {
        theory_->NonCenteredModels(&last_line_models);
      }
      for (int m = 0; m < last_line_models.size(); m++) {
        const ParagraphModel *model = last_line_models[m];
        if (ValidBodyLine(rows_, i, model))
          row.AddBodyLine(model);
      }
    }
 
    // Step Four:
    //   If we're still quite unsure about this line, go through all
    //   models in our theory and see if this row could be the start
    //   of any of our  models.
    if (row.GetLineType() == LT_UNKNOWN ||
        (row.GetLineType() == LT_START && !row.UniqueStartHypothesis())) {
      SetOfModels all_models;
      theory_->NonCenteredModels(&all_models);
      for (int m = 0; m < all_models.size(); m++) {
        if (ValidFirstLine(rows_, i, all_models[m])) {
          row.AddStartLine(all_models[m]);
        }
      }
    }
    // Step Five:
    //   Since we may have updated the hypotheses about this row, we need
    //   to recalculate the Open models for the rest of rows[i + 1, row_end)
    if (row.GetLineType() != LT_UNKNOWN) {
      CalculateOpenModels(i + 1, row_end_);
    }
  }
}
 
// ================ Main Paragraph Detection Algorithm =======================
 
// Find out what ParagraphModels are actually used, and discard any
// that are not.
static void DiscardUnusedModels(const GenericVector<RowScratchRegisters> &rows,
                                ParagraphTheory *theory) {
  SetOfModels used_models;
  for (int i = 0; i < rows.size(); i++) {
    rows[i].StrongHypotheses(&used_models);
  }
  theory->DiscardUnusedModels(used_models);
}
 
// DowngradeWeakestToCrowns:

---

The documentation for this class was generated from the following files:

• /tesseract/src/ccmain/paragraphs_internal.h
• /tesseract/src/ccmain/paragraphs.cpp