tesseract  5.0.0-alpha-619-ge9db
tesseract::LineFinder Class Reference

#include <linefind.h>

Static Public Member Functions

static void FindAndRemoveLines (int resolution, bool debug, Pix *pix, int *vertical_x, int *vertical_y, Pix **pix_music_mask, TabVector_LIST *v_lines, TabVector_LIST *h_lines)
 
static void ConvertBoxaToBlobs (int image_width, int image_height, Boxa **boxes, C_BLOB_LIST *blobs)
 

Detailed Description

The LineFinder class is a simple static function wrapper class that mainly exposes the FindVerticalLines function.

Definition at line 39 of file linefind.h.

Member Function Documentation

◆ ConvertBoxaToBlobs()

void tesseract::LineFinder::ConvertBoxaToBlobs ( int  image_width,
int  image_height,
Boxa **  boxes,
C_BLOB_LIST *  blobs 
)
static

Converts the Boxa array to a list of C_BLOB, getting rid of severely overlapping outlines and those that are children of a bigger one.

The output is a list of C_BLOBs that are owned by the list.

The C_OUTLINEs in the C_BLOBs contain no outline data - just empty bounding boxes. The Boxa is consumed and destroyed.

Definition at line 319 of file linefind.cpp.

320  {
321  C_OUTLINE_LIST outlines;
322  C_OUTLINE_IT ol_it = &outlines;
323  // Iterate the boxes to convert to outlines.
324  int nboxes = boxaGetCount(*boxes);
325  for (int i = 0; i < nboxes; ++i) {
326  l_int32 x, y, width, height;
327  boxaGetBoxGeometry(*boxes, i, &x, &y, &width, &height);
328  // Make a C_OUTLINE from the leptonica box. This is a bit of a hack,
329  // as there is no outline, just a bounding box, but with some very
330  // small changes to coutln.cpp, it works nicely.
331  ICOORD top_left(x, y);
332  ICOORD bot_right(x + width, y + height);
333  CRACKEDGE startpt;
334  startpt.pos = top_left;
335  auto* outline = new C_OUTLINE(&startpt, top_left, bot_right, 0);
336  ol_it.add_after_then_move(outline);
337  }
338  // Use outlines_to_blobs to convert the outlines to blobs and find
339  // overlapping and contained objects. The output list of blobs in the block
340  // has all the bad ones filtered out and deleted.
341  BLOCK block;
342  ICOORD page_tl(0, 0);
343  ICOORD page_br(image_width, image_height);
344  outlines_to_blobs(&block, page_tl, page_br, &outlines);
345  // Transfer the created blobs to the output list.
346  C_BLOB_IT blob_it(blobs);
347  blob_it.add_list_after(block.blob_list());
348  // The boxes aren't needed any more.
349  boxaDestroy(boxes);
350 }

◆ FindAndRemoveLines()

void tesseract::LineFinder::FindAndRemoveLines ( int  resolution,
bool  debug,
Pix *  pix,
int *  vertical_x,
int *  vertical_y,
Pix **  pix_music_mask,
TabVector_LIST *  v_lines,
TabVector_LIST *  h_lines 
)
static

Finds vertical and horizontal line objects in the given pix and removes them.

Uses the given resolution to determine size thresholds instead of any that may be present in the pix.

The output vertical_x and vertical_y contain a sum of the output vectors, thereby giving the mean vertical direction.

If pix_music_mask != nullptr, and music is detected, a mask of the staves and anything that is connected (bars, notes etc.) will be returned in pix_music_mask, the mask subtracted from pix, and the lines will not appear in v_lines or h_lines.

The output vectors are owned by the list and Frozen (cannot refit) by having no boxes, as there is no need to refit or merge separator lines.

The detected lines are removed from the pix.

Definition at line 243 of file linefind.cpp.

247  {
248  if (pix == nullptr || vertical_x == nullptr || vertical_y == nullptr) {
249  tprintf("Error in parameters for LineFinder::FindAndRemoveLines\n");
250  return;
251  }
252  Pix* pix_vline = nullptr;
253  Pix* pix_non_vline = nullptr;
254  Pix* pix_hline = nullptr;
255  Pix* pix_non_hline = nullptr;
256  Pix* pix_intersections = nullptr;
257  Pixa* pixa_display = debug ? pixaCreate(0) : nullptr;
258  GetLineMasks(resolution, pix, &pix_vline, &pix_non_vline, &pix_hline,
259  &pix_non_hline, &pix_intersections, pix_music_mask,
260  pixa_display);
261  // Find lines, convert to TabVector_LIST and remove those that are used.
262  FindAndRemoveVLines(resolution, pix_intersections, vertical_x, vertical_y,
263  &pix_vline, pix_non_vline, pix, v_lines);
264  if (pix_hline != nullptr) {
265  // Recompute intersections and re-filter false positive h-lines.
266  if (pix_vline != nullptr)
267  pixAnd(pix_intersections, pix_vline, pix_hline);
268  else
269  pixDestroy(&pix_intersections);
270  if (!FilterFalsePositives(resolution, pix_non_hline, pix_intersections,
271  pix_hline)) {
272  pixDestroy(&pix_hline);
273  }
274  }
275  FindAndRemoveHLines(resolution, pix_intersections, *vertical_x, *vertical_y,
276  &pix_hline, pix_non_hline, pix, h_lines);
277  if (pixa_display != nullptr && pix_vline != nullptr)
278  pixaAddPix(pixa_display, pix_vline, L_CLONE);
279  if (pixa_display != nullptr && pix_hline != nullptr)
280  pixaAddPix(pixa_display, pix_hline, L_CLONE);
281  if (pix_vline != nullptr && pix_hline != nullptr) {
282  // Remove joins (intersections) where lines cross, and the residue.
283  // Recalculate the intersections, since some lines have been deleted.
284  pixAnd(pix_intersections, pix_vline, pix_hline);
285  // Fatten up the intersections and seed-fill to get the intersection
286  // residue.
287  Pix* pix_join_residue = pixDilateBrick(nullptr, pix_intersections, 5, 5);
288  pixSeedfillBinary(pix_join_residue, pix_join_residue, pix, 8);
289  // Now remove the intersection residue.
290  pixSubtract(pix, pix, pix_join_residue);
291  pixDestroy(&pix_join_residue);
292  }
293  // Remove any detected music.
294  if (pix_music_mask != nullptr && *pix_music_mask != nullptr) {
295  if (pixa_display != nullptr)
296  pixaAddPix(pixa_display, *pix_music_mask, L_CLONE);
297  pixSubtract(pix, pix, *pix_music_mask);
298  }
299  if (pixa_display != nullptr)
300  pixaAddPix(pixa_display, pix, L_CLONE);
301 
302  pixDestroy(&pix_vline);
303  pixDestroy(&pix_non_vline);
304  pixDestroy(&pix_hline);
305  pixDestroy(&pix_non_hline);
306  pixDestroy(&pix_intersections);
307  if (pixa_display != nullptr) {
308  pixaConvertToPdf(pixa_display, resolution, 1.0f, 0, 0, "LineFinding",
309  "vhlinefinding.pdf");
310  pixaDestroy(&pixa_display);
311  }
312 }

The documentation for this class was generated from the following files:
ICOORD
integer coordinate
Definition: points.h:30
C_OUTLINE
class DLLSYM C_OUTLINE
Definition: coutln.h:67
BLOCK
Definition: ocrblock.h:28
CRACKEDGE::pos
ICOORD pos
Definition: crakedge.h:45
outlines_to_blobs
void outlines_to_blobs(BLOCK *block, ICOORD bleft, ICOORD tright, C_OUTLINE_LIST *outlines)
Definition: edgblob.cpp:349
CRACKEDGE
Definition: crakedge.h:25
BLOCK::blob_list
C_BLOB_LIST * blob_list()
get blobs
Definition: ocrblock.h:127
tprintf
DLLSYM void tprintf(const char *format,...)
Definition: tprintf.cpp:34