void draw_meanlines( //draw a block
TO_BLOCK *block, //block to draw
float gradient, //gradients of lines
inT32 left, //edge of block
ScrollView::Color colour, //colour to draw in
FCOORD rotation //rotation for line
) {
FCOORD plot_pt; //point to plot
//rows
TO_ROW_IT row_it = block->get_rows();
TO_ROW *row; //current row
BLOBNBOX_IT blob_it; //blobs
float right; //end of row
to_win->Pen(colour);
for (row_it.mark_cycle_pt(); !row_it.cycled_list(); row_it.forward()) {
row = row_it.data();
blob_it.set_to_list(row->blob_list());
blob_it.move_to_last();
right = blob_it.data()->bounding_box().right();
plot_pt =
FCOORD((float) left,
gradient * left + row->parallel_c() + row->xheight);
plot_pt.rotate(rotation);
to_win->SetCursor(plot_pt.x(), plot_pt.y());
plot_pt =
FCOORD((float) right,
gradient * right + row->parallel_c() + row->xheight);
plot_pt.rotate(rotation);
to_win->DrawTo(plot_pt.x(), plot_pt.y());
}
}
// Fits splines to the textlines, or creates fake QSPLINES from the straight
// baselines that are already on the TO_ROWs.
// As a side-effect, computes the xheights of the rows and the block.
// Although x-height estimation is conceptually separate, it is part of
// detecting perspective distortion and therefore baseline fitting.
void BaselineBlock::FitBaselineSplines(bool enable_splines,
bool show_final_rows,
Textord* textord) {
double gradient = tan(skew_angle_);
FCOORD rotation(1.0f, 0.0f);
if (enable_splines) {
textord->make_spline_rows(block_, gradient, show_final_rows);
} else {
// Make a fake spline from the existing line.
TBOX block_box= block_->block->bounding_box();
TO_ROW_IT row_it = block_->get_rows();
for (row_it.mark_cycle_pt(); !row_it.cycled_list(); row_it.forward()) {
TO_ROW* row = row_it.data();
inT32 xstarts[2] = { block_box.left(), block_box.right() };
double coeffs[3] = { 0.0, row->line_m(), row->line_c() };
row->baseline = QSPLINE(1, xstarts, coeffs);
textord->compute_row_xheight(row, block_->block->classify_rotation(),
row->line_m(), block_->line_size);
}
}
textord->compute_block_xheight(block_, gradient);
block_->block->set_xheight(block_->xheight);
if (textord_restore_underlines) // fix underlines
restore_underlined_blobs(block_);
}
void make_single_word(bool one_blob, TO_ROW_LIST *rows, ROW_LIST* real_rows) {
TO_ROW_IT to_row_it(rows);
TO_ROW* row = to_row_it.data();
// The blobs have to come out of the BLOBNBOX into the C_BLOB_LIST ready
// to create the word.
C_BLOB_LIST cblobs;
C_BLOB_IT cblob_it(&cblobs);
BLOBNBOX_IT box_it(row->blob_list());
for (;!box_it.empty(); box_it.forward()) {
BLOBNBOX* bblob= box_it.extract();
if (bblob->joined_to_prev() || (one_blob && !cblob_it.empty())) {
if (bblob->cblob() != NULL) {
C_OUTLINE_IT cout_it(cblob_it.data()->out_list());
cout_it.move_to_last();
cout_it.add_list_after(bblob->cblob()->out_list());
delete bblob->cblob();
}
} else {
if (bblob->cblob() != NULL)
cblob_it.add_after_then_move(bblob->cblob());
delete bblob;
}
}
// Convert the TO_ROW to a ROW.
ROW* real_row = new ROW(row, static_cast<inT16>(row->kern_size),
static_cast<inT16>(row->space_size));
WERD_IT word_it(real_row->word_list());
WERD* word = new WERD(&cblobs, 0, NULL);
word->set_flag(W_BOL, TRUE);
word->set_flag(W_EOL, TRUE);
word_it.add_after_then_move(word);
ROW_IT row_it(real_rows);
row_it.add_after_then_move(real_row);
}
//yangjing01 modified :
bool TAL_make_single_word(bool one_blob, TO_ROW_LIST* rows, ROW_LIST* real_rows)
{
TO_ROW_IT to_row_it(rows);
ROW_IT row_it(real_rows);
//to_real_row is the real row information of single row or single char mode
TO_ROW* real_to_row = NULL;
float row_max_height = 0.0;
for (to_row_it.mark_cycle_pt();
!to_row_it.cycled_list(); to_row_it.forward()){
TO_ROW* row = to_row_it.data();
float row_min_y = row->min_y();
float row_max_y = row->max_y();
float row_height = abs(row_max_y - row_min_y);
if (real_to_row == NULL
|| row_height > row_max_height
|| fabs(row_height - row_max_height) < 1.0f){
row_max_height = row_height;
real_to_row = row;
}
}
if (real_to_row == NULL){
return false;
}
C_BLOB_LIST cblobs;
C_BLOB_IT cblob_it(&cblobs);
BLOBNBOX_IT box_it(real_to_row->blob_list());
for (; !box_it.empty(); box_it.forward()){
BLOBNBOX* bblob = box_it.extract();
if (bblob->joined_to_prev() || (one_blob && !cblob_it.empty())) {
if (bblob->cblob() != NULL){
C_OUTLINE_IT cout_it(cblob_it.data()->out_list());
cout_it.move_to_last();
cout_it.add_list_after(bblob->cblob()->out_list());
delete bblob->cblob();
}
}
else {
if (bblob->cblob() != NULL)
cblob_it.add_after_then_move(bblob->cblob());
}
delete bblob;
}
// Convert the TO_ROW to a ROW.
ROW* real_row = new ROW(real_to_row, static_cast<inT16>(real_to_row->kern_size),
static_cast<inT16>(real_to_row->space_size));
WERD_IT word_it(real_row->word_list());
WERD* word = new WERD(&cblobs, 0, NULL);
word->set_flag(W_BOL, TRUE);
word->set_flag(W_EOL, TRUE);
word->set_flag(W_DONT_CHOP, one_blob);
word_it.add_after_then_move(word);
row_it.add_after_then_move(real_row);
return true;
}
void make_real_words(
tesseract::Textord *textord,
TO_BLOCK *block, //block to do
FCOORD rotation //for drawing
) {
TO_ROW *row; //current row
TO_ROW_IT row_it = block->get_rows ();
ROW *real_row = NULL; //output row
ROW_IT real_row_it = block->block->row_list ();
if (row_it.empty ())
return; //empty block
for (row_it.mark_cycle_pt (); !row_it.cycled_list (); row_it.forward ()) {
row = row_it.data ();
if (row->blob_list ()->empty () && !row->rep_words.empty ()) {
real_row = make_rep_words (row, block);
} else if (!row->blob_list()->empty()) {
// In a fixed pitch document, some lines may be detected as fixed pitch
// while others don't, and will go through different path.
// For non-space delimited language like CJK, fixed pitch chop always
// leave the entire line as one word. We can force consistent chopping
// with force_make_prop_words flag.
POLY_BLOCK* pb = block->block->poly_block();
if (textord_chopper_test) {
real_row = textord->make_blob_words (row, rotation);
} else if (textord_force_make_prop_words ||
(pb != NULL && !pb->IsText()) ||
row->pitch_decision == PITCH_DEF_PROP ||
row->pitch_decision == PITCH_CORR_PROP) {
real_row = textord->make_prop_words (row, rotation);
} else if (row->pitch_decision == PITCH_DEF_FIXED ||
row->pitch_decision == PITCH_CORR_FIXED) {
real_row = fixed_pitch_words (row, rotation);
} else {
ASSERT_HOST(FALSE);
}
}
if (real_row != NULL) {
//put row in block
real_row_it.add_after_then_move (real_row);
}
}
block->block->set_stats (block->fixed_pitch == 0, (inT16) block->kern_size,
(inT16) block->space_size,
(inT16) block->fixed_pitch);
block->block->check_pitch ();
}
void set_row_spaces( //find space sizes
TO_BLOCK *block, //block to do
FCOORD rotation, //for drawing
BOOL8 testing_on //correct orientation
) {
inT32 maxwidth; //of widest space
TO_ROW *row; //current row
TO_ROW_IT row_it = block->get_rows ();
if (row_it.empty ())
return; //empty block
maxwidth = (inT32) ceil (block->xheight * textord_words_maxspace);
for (row_it.mark_cycle_pt (); !row_it.cycled_list (); row_it.forward ()) {
row = row_it.data ();
if (row->fixed_pitch == 0) {
// if (!textord_test_mode
// && row_words(block,row,maxwidth,rotation,testing_on)==0
// || textord_test_mode
// && row_words2(block,row,maxwidth,rotation,testing_on)==0)
// {
row->min_space =
(inT32) ceil (row->pr_space -
(row->pr_space -
row->pr_nonsp) * textord_words_definite_spread);
row->max_nonspace =
(inT32) floor (row->pr_nonsp +
(row->pr_space -
row->pr_nonsp) * textord_words_definite_spread);
if (testing_on && textord_show_initial_words) {
tprintf ("Assigning defaults %d non, %d space to row at %g\n",
row->max_nonspace, row->min_space, row->intercept ());
}
row->space_threshold = (row->max_nonspace + row->min_space) / 2;
row->space_size = row->pr_space;
row->kern_size = row->pr_nonsp;
// }
}
#ifndef GRAPHICS_DISABLED
if (textord_show_initial_words && testing_on) {
plot_word_decisions (to_win, (inT16) row->fixed_pitch, row);
}
#endif
}
}
void set_row_spaces( //find space sizes
TO_BLOCK* block, //block to do
FCOORD rotation, //for drawing
bool testing_on //correct orientation
) {
TO_ROW *row; //current row
TO_ROW_IT row_it = block->get_rows ();
if (row_it.empty ())
return; //empty block
for (row_it.mark_cycle_pt (); !row_it.cycled_list (); row_it.forward ()) {
row = row_it.data ();
if (row->fixed_pitch == 0) {
row->min_space =
static_cast<int32_t>(ceil (row->pr_space -
(row->pr_space -
row->pr_nonsp) * textord_words_definite_spread));
row->max_nonspace =
static_cast<int32_t>(floor (row->pr_nonsp +
(row->pr_space -
row->pr_nonsp) * textord_words_definite_spread));
if (testing_on && textord_show_initial_words) {
tprintf ("Assigning defaults %d non, %d space to row at %g\n",
row->max_nonspace, row->min_space, row->intercept ());
}
row->space_threshold = (row->max_nonspace + row->min_space) / 2;
row->space_size = row->pr_space;
row->kern_size = row->pr_nonsp;
}
#ifndef GRAPHICS_DISABLED
if (textord_show_initial_words && testing_on) {
plot_word_decisions (to_win, static_cast<int16_t>(row->fixed_pitch), row);
}
#endif
}
}
void restore_underlined_blobs( //get chop points
TO_BLOCK *block //block to do
) {
int16_t chop_coord; //chop boundary
TBOX blob_box; //of underline
BLOBNBOX *u_line; //underline bit
TO_ROW *row; //best row for blob
ICOORDELT_LIST chop_cells; //blobs to cut out
//real underlines
BLOBNBOX_LIST residual_underlines;
C_OUTLINE_LIST left_coutlines;
C_OUTLINE_LIST right_coutlines;
ICOORDELT_IT cell_it = &chop_cells;
//under lines
BLOBNBOX_IT under_it = &block->underlines;
BLOBNBOX_IT ru_it = &residual_underlines;
if (block->get_rows()->empty())
return; // Don't crash if there are no rows.
for (under_it.mark_cycle_pt (); !under_it.cycled_list ();
under_it.forward ()) {
u_line = under_it.extract ();
blob_box = u_line->bounding_box ();
row = most_overlapping_row (block->get_rows (), u_line);
if (row == nullptr)
return; // Don't crash if there is no row.
find_underlined_blobs (u_line, &row->baseline, row->xheight,
row->xheight * textord_underline_offset,
&chop_cells);
cell_it.set_to_list (&chop_cells);
for (cell_it.mark_cycle_pt (); !cell_it.cycled_list ();
cell_it.forward ()) {
chop_coord = cell_it.data ()->x ();
if (cell_it.data ()->y () - chop_coord > textord_fp_chop_error + 1) {
split_to_blob (u_line, chop_coord,
textord_fp_chop_error + 0.5,
&left_coutlines,
&right_coutlines);
if (!left_coutlines.empty()) {
ru_it.add_after_then_move(new BLOBNBOX(new C_BLOB(&left_coutlines)));
}
chop_coord = cell_it.data ()->y ();
split_to_blob(nullptr, chop_coord, textord_fp_chop_error + 0.5,
&left_coutlines, &right_coutlines);
if (!left_coutlines.empty()) {
row->insert_blob(new BLOBNBOX(new C_BLOB(&left_coutlines)));
}
u_line = nullptr; //no more blobs to add
}
delete cell_it.extract();
}
if (!right_coutlines.empty ()) {
split_to_blob(nullptr, blob_box.right(), textord_fp_chop_error + 0.5,
&left_coutlines, &right_coutlines);
if (!left_coutlines.empty())
ru_it.add_after_then_move(new BLOBNBOX(new C_BLOB(&left_coutlines)));
}
if (u_line != nullptr) {
delete u_line->cblob();
delete u_line;
}
}
if (!ru_it.empty()) {
ru_it.move_to_first();
for (ru_it.mark_cycle_pt(); !ru_it.cycled_list(); ru_it.forward()) {
under_it.add_after_then_move(ru_it.extract());
}
}
}
GAPMAP::GAPMAP( //Constructor
TO_BLOCK *block //block
) {
TO_ROW_IT row_it; //row iterator
TO_ROW *row; //current row
BLOBNBOX_IT blob_it; //iterator
TBOX blob_box;
TBOX prev_blob_box;
inT16 gap_width;
inT16 start_of_row;
inT16 end_of_row;
STATS xht_stats (0, 128);
inT16 min_quantum;
inT16 max_quantum;
inT16 i;
row_it.set_to_list (block->get_rows ());
/*
Find left and right extremes and bucket size
*/
map = NULL;
min_left = MAX_INT16;
max_right = -MAX_INT16;
total_rows = 0;
any_tabs = FALSE;
for (row_it.mark_cycle_pt (); !row_it.cycled_list (); row_it.forward ()) {
row = row_it.data ();
if (!row->blob_list ()->empty ()) {
total_rows++;
xht_stats.add ((inT16) floor (row->xheight + 0.5), 1);
blob_it.set_to_list (row->blob_list ());
start_of_row = blob_it.data ()->bounding_box ().left ();
end_of_row = blob_it.data_relative (-1)->bounding_box ().right ();
if (min_left > start_of_row)
min_left = start_of_row;
if (max_right < end_of_row)
max_right = end_of_row;
}
}
if ((total_rows < 3) || (min_left >= max_right)) {
total_rows = 0;
min_left = max_right = 0;
return;
}
bucket_size = (inT16) floor (xht_stats.median () + 0.5) / 2;
map_max = (max_right - min_left) / bucket_size;
map = (inT16 *) alloc_mem ((map_max + 1) * sizeof (inT16));
for (i = 0; i <= map_max; i++)
map[i] = 0;
for (row_it.mark_cycle_pt (); !row_it.cycled_list (); row_it.forward ()) {
row = row_it.data ();
if (!row->blob_list ()->empty ()) {
blob_it.set_to_list (row->blob_list ());
blob_it.mark_cycle_pt ();
blob_box = box_next (&blob_it);
prev_blob_box = blob_box;
if (gapmap_use_ends) {
/* Leading space */
gap_width = blob_box.left () - min_left;
if ((gap_width > gapmap_big_gaps * row->xheight)
&& gap_width > 2) {
max_quantum = (blob_box.left () - min_left) / bucket_size;
for (i = 0; i <= max_quantum; i++)
map[i]++;
}
}
while (!blob_it.cycled_list ()) {
blob_box = box_next (&blob_it);
gap_width = blob_box.left () - prev_blob_box.right ();
if ((gap_width > gapmap_big_gaps * row->xheight)
&& gap_width > 2) {
min_quantum =
(prev_blob_box.right () - min_left) / bucket_size;
max_quantum = (blob_box.left () - min_left) / bucket_size;
for (i = min_quantum; i <= max_quantum; i++)
map[i]++;
}
prev_blob_box = blob_box;
}
if (gapmap_use_ends) {
/* Trailing space */
gap_width = max_right - prev_blob_box.right ();
if ((gap_width > gapmap_big_gaps * row->xheight)
&& gap_width > 2) {
min_quantum =
(prev_blob_box.right () - min_left) / bucket_size;
for (i = min_quantum; i <= map_max; i++)
map[i]++;
}
}
}
}
for (i = 0; i <= map_max; i++) {
if (map[i] > total_rows / 2) {
if (gapmap_no_isolated_quanta &&
(((i == 0) &&
(map[i + 1] <= total_rows / 2)) ||
((i == map_max) &&
(map[i - 1] <= total_rows / 2)) ||
((i > 0) &&
(i < map_max) &&
(map[i - 1] <= total_rows / 2) &&
(map[i + 1] <= total_rows / 2)))) {
map[i] = 0; //prevent isolated quantum
}
else
any_tabs = TRUE;
}
}
if (gapmap_debug && any_tabs)
tprintf ("Table found\n");
}