#include <devanagari_processing.h>

Public Types
enum	SplitStrategy { NO_SPLIT = 0, MINIMAL_SPLIT, MAXIMAL_SPLIT }
Public Member Functions
	ShiroRekhaSplitter ()
virtual	~ShiroRekhaSplitter ()
bool	Split (bool split_for_pageseg)
void	Clear ()
void	RefreshSegmentationWithNewBlobs (C_BLOB_LIST *new_blobs)
bool	HasDifferentSplitStrategies () const
void	set_segmentation_block_list (BLOCK_LIST *block_list)
void	set_global_xheight (int xheight)
void	set_perform_close (bool perform)
Pix *	splitted_image ()
void	set_orig_pix (Pix *pix)
Pix *	orig_pix ()
SplitStrategy	ocr_split_strategy () const
void	set_ocr_split_strategy (SplitStrategy strategy)
SplitStrategy	pageseg_split_strategy () const
void	set_pageseg_split_strategy (SplitStrategy strategy)
BLOCK_LIST *	segmentation_block_list ()
void	DumpDebugImage (const char *filename) const
Static Public Member Functions
static int	GetModeHeight (Pix *pix)
Static Public Attributes
static const int	kUnspecifiedXheight = -1

Detailed Description

Definition at line 66 of file devanagari_processing.h.

Member Enumeration Documentation

enum tesseract::ShiroRekhaSplitter::SplitStrategy

Enumerator:

NO_SPLIT
MINIMAL_SPLIT
MAXIMAL_SPLIT

Definition at line 68 of file devanagari_processing.h.

                     {
    NO_SPLIT = 0,   // No splitting is performed for the phase.
    MINIMAL_SPLIT,  // Blobs are split minimally.
    MAXIMAL_SPLIT   // Blobs are split maximally.
  };

Constructor & Destructor Documentation

tesseract::ShiroRekhaSplitter::ShiroRekhaSplitter ( )

Definition at line 37 of file devanagari_processing.cpp.

                                       {
  orig_pix_ = NULL;
  segmentation_block_list_ = NULL;
  splitted_image_ = NULL;
  global_xheight_ = kUnspecifiedXheight;
  perform_close_ = false;
  debug_image_ = NULL;
  pageseg_split_strategy_ = NO_SPLIT;
  ocr_split_strategy_ = NO_SPLIT;
}

tesseract::ShiroRekhaSplitter::~ShiroRekhaSplitter ( ) [virtual]

Definition at line 48 of file devanagari_processing.cpp.

                                        {
  Clear();
}

Member Function Documentation

void tesseract::ShiroRekhaSplitter::Clear ( )

Definition at line 52 of file devanagari_processing.cpp.

                               {
  pixDestroy(&orig_pix_);
  pixDestroy(&splitted_image_);
  pageseg_split_strategy_ = NO_SPLIT;
  ocr_split_strategy_ = NO_SPLIT;
  pixDestroy(&debug_image_);
  segmentation_block_list_ = NULL;
  global_xheight_ = kUnspecifiedXheight;
  perform_close_ = false;
}

void tesseract::ShiroRekhaSplitter::DumpDebugImage ( const char * filename ) const

Definition at line 64 of file devanagari_processing.cpp.

                                                                  {
  pixWrite(filename, debug_image_, IFF_PNG);
}

int tesseract::ShiroRekhaSplitter::GetModeHeight ( Pix * pix ) [static]

Definition at line 409 of file devanagari_processing.cpp.

                                              {
  Boxa* boxa = pixConnComp(pix, NULL, 8);
  STATS heights(0, pixGetHeight(pix));
  heights.clear();
  for (int i = 0; i < boxaGetCount(boxa); ++i) {
    Box* box = boxaGetBox(boxa, i, L_CLONE);
    if (box->h >= 3 || box->w >= 3) {
      heights.add(box->h, 1);
    }
    boxDestroy(&box);
  }
  boxaDestroy(&boxa);
  return heights.mode();
}

bool tesseract::ShiroRekhaSplitter::HasDifferentSplitStrategies ( ) const [inline]

Definition at line 92 of file devanagari_processing.h.

                                           {
    return pageseg_split_strategy_ != ocr_split_strategy_;
  }

SplitStrategy tesseract::ShiroRekhaSplitter::ocr_split_strategy ( ) const [inline]

Definition at line 129 of file devanagari_processing.h.

                                           {
    return ocr_split_strategy_;
  }

Pix* tesseract::ShiroRekhaSplitter::orig_pix ( ) [inline]

Definition at line 125 of file devanagari_processing.h.

                  {
    return orig_pix_;
  }

SplitStrategy tesseract::ShiroRekhaSplitter::pageseg_split_strategy ( ) const [inline]

Definition at line 137 of file devanagari_processing.h.

                                               {
    return pageseg_split_strategy_;
  }

void tesseract::ShiroRekhaSplitter::RefreshSegmentationWithNewBlobs ( C_BLOB_LIST * new_blobs )

Definition at line 355 of file devanagari_processing.cpp.

                            {
  // The segmentation block list must have been specified.
  ASSERT_HOST(segmentation_block_list_);
  if (devanagari_split_debuglevel > 0) {
    tprintf("Before refreshing blobs:\n");
    PrintSegmentationStats(segmentation_block_list_);
    tprintf("New Blobs found: %d\n", new_blobs->length());
  }

  C_BLOB_LIST not_found_blobs;
  RefreshWordBlobsFromNewBlobs(segmentation_block_list_,
                               new_blobs,
                               ((devanagari_split_debugimage && debug_image_) ?
                                &not_found_blobs : NULL));

  if (devanagari_split_debuglevel > 0) {
    tprintf("After refreshing blobs:\n");
    PrintSegmentationStats(segmentation_block_list_);
  }
  if (devanagari_split_debugimage && debug_image_) {
    // Plot out the original blobs for which no match was found in the new
    // all_blobs list.
    C_BLOB_IT not_found_it(&not_found_blobs);
    for (not_found_it.mark_cycle_pt(); !not_found_it.cycled_list();
         not_found_it.forward()) {
      C_BLOB* not_found = not_found_it.data();
      TBOX not_found_box = not_found->bounding_box();
      Box* box_to_plot = GetBoxForTBOX(not_found_box);
      pixRenderBoxArb(debug_image_, box_to_plot, 1, 255, 0, 255);
      boxDestroy(&box_to_plot);
    }

    // Plot out the blobs unused from all blobs.
    C_BLOB_IT all_blobs_it(new_blobs);
    for (all_blobs_it.mark_cycle_pt(); !all_blobs_it.cycled_list();
         all_blobs_it.forward()) {
      C_BLOB* a_blob = all_blobs_it.data();
      Box* box_to_plot = GetBoxForTBOX(a_blob->bounding_box());
      pixRenderBoxArb(debug_image_, box_to_plot, 3, 0, 127, 0);
      boxDestroy(&box_to_plot);
    }
  }
}

BLOCK_LIST* tesseract::ShiroRekhaSplitter::segmentation_block_list ( ) [inline]

Definition at line 145 of file devanagari_processing.h.

                                        {
    return segmentation_block_list_;
  }

void tesseract::ShiroRekhaSplitter::set_global_xheight ( int xheight ) [inline]

Definition at line 105 of file devanagari_processing.h.

                                       {
    global_xheight_ = xheight;
  }

void tesseract::ShiroRekhaSplitter::set_ocr_split_strategy ( SplitStrategy strategy ) [inline]

Definition at line 133 of file devanagari_processing.h.

                                                      {
    ocr_split_strategy_ = strategy;
  }

void tesseract::ShiroRekhaSplitter::set_orig_pix ( Pix * pix )

Definition at line 69 of file devanagari_processing.cpp.

                                              {
  if (orig_pix_) {
    pixDestroy(&orig_pix_);
  }
  orig_pix_ = pixClone(pix);
}

void tesseract::ShiroRekhaSplitter::set_pageseg_split_strategy ( SplitStrategy strategy ) [inline]

Definition at line 141 of file devanagari_processing.h.

                                                          {
    pageseg_split_strategy_ = strategy;
  }

void tesseract::ShiroRekhaSplitter::set_perform_close ( bool perform ) [inline]

Definition at line 109 of file devanagari_processing.h.

                                       {
    perform_close_ = perform;
  }

void tesseract::ShiroRekhaSplitter::set_segmentation_block_list ( BLOCK_LIST * block_list ) [inline]

Definition at line 99 of file devanagari_processing.h.

                                                           {
    segmentation_block_list_ = block_list;
  }

bool tesseract::ShiroRekhaSplitter::Split ( bool split_for_pageseg )

Definition at line 81 of file devanagari_processing.cpp.

                                                     {
  SplitStrategy split_strategy = split_for_pageseg ? pageseg_split_strategy_ :
      ocr_split_strategy_;
  if (split_strategy == NO_SPLIT) {
    return false;  // Nothing to do.
  }
  ASSERT_HOST(split_strategy == MINIMAL_SPLIT ||
              split_strategy == MAXIMAL_SPLIT);
  ASSERT_HOST(orig_pix_);
  if (devanagari_split_debuglevel > 0) {
    tprintf("Splitting shiro-rekha ...\n");
    tprintf("Split strategy = %s\n",
            split_strategy == MINIMAL_SPLIT ? "Minimal" : "Maximal");
    tprintf("Initial pageseg available = %s\n",
            segmentation_block_list_ ? "yes" : "no");
  }
  // Create a copy of original image to store the splitting output.
  pixDestroy(&splitted_image_);
  splitted_image_ = pixCopy(NULL, orig_pix_);

  // Initialize debug image if required.
  if (devanagari_split_debugimage) {
    pixDestroy(&debug_image_);
    debug_image_ = pixConvertTo32(orig_pix_);
  }

  // Determine all connected components in the input image. A close operation
  // may be required prior to this, depending on the current settings.
  Pix* pix_for_ccs = pixClone(orig_pix_);
  if (perform_close_ && global_xheight_ != kUnspecifiedXheight &&
      !segmentation_block_list_) {
    if (devanagari_split_debuglevel > 0) {
      tprintf("Performing a global close operation..\n");
    }
    // A global measure is available for xheight, but no local information
    // exists.
    pixDestroy(&pix_for_ccs);
    pix_for_ccs = pixCopy(NULL, orig_pix_);
    PerformClose(pix_for_ccs, global_xheight_);
  }
  Pixa* ccs;
  Boxa* tmp_boxa = pixConnComp(pix_for_ccs, &ccs, 8);
  boxaDestroy(&tmp_boxa);
  pixDestroy(&pix_for_ccs);

  // Iterate over all connected components. Get their bounding boxes and clip
  // out the image regions corresponding to these boxes from the original image.
  // Conditionally run splitting on each of them.
  Boxa* regions_to_clear = boxaCreate(0);
  for (int i = 0; i < pixaGetCount(ccs); ++i) {
    Box* box = ccs->boxa->box[i];
    Pix* word_pix = pixClipRectangle(orig_pix_, box, NULL);
    ASSERT_HOST(word_pix);
    int xheight = GetXheightForCC(box);
    if (xheight == kUnspecifiedXheight && segmentation_block_list_ &&
        devanagari_split_debugimage) {
      pixRenderBoxArb(debug_image_, box, 1, 255, 0, 0);
    }
    // If some xheight measure is available, attempt to pre-eliminate small
    // blobs from the shiro-rekha process. This is primarily to save the CCs
    // corresponding to punctuation marks/small dots etc which are part of
    // larger graphemes.
    if (xheight == kUnspecifiedXheight ||
        (box->w > xheight / 3 && box->h > xheight / 2)) {
      SplitWordShiroRekha(split_strategy, word_pix, xheight,
                          box->x, box->y, regions_to_clear);
    } else if (devanagari_split_debuglevel > 0) {
      tprintf("CC dropped from splitting: %d,%d (%d, %d)\n",
              box->x, box->y, box->w, box->h);
    }
    pixDestroy(&word_pix);
  }
  // Actually clear the boxes now.
  for (int i = 0; i < boxaGetCount(regions_to_clear); ++i) {
    Box* box = boxaGetBox(regions_to_clear, i, L_CLONE);
    pixClearInRect(splitted_image_, box);
    boxDestroy(&box);
  }
  boxaDestroy(&regions_to_clear);
  pixaDestroy(&ccs);
  if (devanagari_split_debugimage) {
    DumpDebugImage(split_for_pageseg ? "pageseg_split_debug.png" :
                   "ocr_split_debug.png");
  }
  return true;
}

Pix* tesseract::ShiroRekhaSplitter::splitted_image ( ) [inline]

Definition at line 116 of file devanagari_processing.h.

                        {
    return splitted_image_;
  }

Member Data Documentation

const int tesseract::ShiroRekhaSplitter::kUnspecifiedXheight = -1 [static]

Definition at line 103 of file devanagari_processing.h.

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

tesseract-ocr/textord/devanagari_processing.h
tesseract-ocr/textord/devanagari_processing.cpp

Public Types

Public Member Functions

Static Public Member Functions

Static Public Attributes

Detailed Description

Member Enumeration Documentation

Constructor & Destructor Documentation

Member Function Documentation

Member Data Documentation