void plot_fp_cells2( //draw words
ScrollView *win, //window tro draw in
ScrollView::Color colour, //colour of lines
TO_ROW *row, //for location
FPSEGPT_LIST *seg_list //segments to plot
) {
TBOX word_box; //bounding box
FPSEGPT_IT seg_it = seg_list;
//blobs in row
BLOBNBOX_IT blob_it = row->blob_list();
FPSEGPT *segpt; //current point
word_box = blob_it.data()->bounding_box();
for (blob_it.mark_cycle_pt(); !blob_it.cycled_list();)
word_box += box_next(&blob_it);
for (seg_it.mark_cycle_pt(); !seg_it.cycled_list(); seg_it.forward()) {
segpt = seg_it.data();
if (segpt->faked) {
colour = ScrollView::WHITE;
win->Pen(colour);
}
else {
win->Pen(colour);
}
win->Line(segpt->position(), word_box.bottom(), segpt->position(), word_box.top());
}
}
void make_illegal_segment( //find segmentation
FPSEGPT_LIST *prev_list, //previous segments
TBOX blob_box, //bounding box
BLOBNBOX_IT blob_it, //iterator
int16_t region_index, //number of segment
int16_t pitch, //pitch estimate
int16_t pitch_error, //tolerance
FPSEGPT_LIST *seg_list //output list
) {
int16_t x; //current coord
int16_t min_x = 0; //in this region
int16_t max_x = 0;
int16_t offset; //dist to edge
FPSEGPT *segpt; //segment point
FPSEGPT *prevpt; //previous point
float best_cost; //best path
FPSEGPT_IT segpt_it = seg_list;//iterator
//previous points
FPSEGPT_IT prevpt_it = prev_list;
best_cost = FLT_MAX;
for (prevpt_it.mark_cycle_pt (); !prevpt_it.cycled_list ();
prevpt_it.forward ()) {
prevpt = prevpt_it.data ();
if (prevpt->cost_function () < best_cost) {
//find least
best_cost = prevpt->cost_function ();
min_x = prevpt->position ();
max_x = min_x; //limits on coords
}
else if (prevpt->cost_function () == best_cost) {
max_x = prevpt->position ();
}
}
min_x += pitch - pitch_error;
max_x += pitch + pitch_error;
for (x = min_x; x <= max_x; x++) {
while (x > blob_box.right ()) {
blob_box = box_next (&blob_it);
}
offset = x - blob_box.left ();
if (blob_box.right () - x < offset)
offset = blob_box.right () - x;
segpt = new FPSEGPT (x, FALSE, offset,
region_index, pitch, pitch_error, prev_list);
if (segpt->previous () != nullptr) {
ASSERT_HOST (offset >= 0);
fprintf (stderr, "made fake at %d\n", x);
//make one up
segpt_it.add_after_then_move (segpt);
segpt->faked = TRUE;
segpt->fake_count++;
}
else
delete segpt;
}
}
void plot_row_cells( //draw words
ScrollView *win, //window tro draw in
ScrollView::Color colour, //colour of lines
TO_ROW *row, //for location
float xshift, //amount of shift
ICOORDELT_LIST *cells //cells to draw
) {
TBOX word_box; //bounding box
ICOORDELT_IT cell_it = cells;
//blobs in row
BLOBNBOX_IT blob_it = row->blob_list();
ICOORDELT *cell; //current cell
word_box = blob_it.data()->bounding_box();
for (blob_it.mark_cycle_pt(); !blob_it.cycled_list();)
word_box += box_next(&blob_it);
win->Pen(colour);
for (cell_it.mark_cycle_pt(); !cell_it.cycled_list(); cell_it.forward()) {
cell = cell_it.data();
win->Line(cell->x() + xshift, word_box.bottom(), cell->x() + xshift, word_box.top());
}
}
void plot_fp_cells( //draw words
ScrollView *win, //window tro draw in
ScrollView::Color colour, //colour of lines
BLOBNBOX_IT *blob_it, //blobs
inT16 pitch, //of block
inT16 blob_count, //no of real blobs
STATS *projection, //vertical
inT16 projection_left, //edges //scale factor
inT16 projection_right,
float projection_scale) {
inT16 occupation; //occupied cells
TBOX word_box; //bounding box
FPSEGPT_LIST seg_list; //list of cuts
FPSEGPT_IT seg_it;
FPSEGPT *segpt; //current point
if (pitsync_linear_version)
check_pitch_sync2(blob_it, blob_count, pitch, 2, projection,
projection_left, projection_right,
projection_scale, occupation, &seg_list, 0, 0);
else
check_pitch_sync(blob_it, blob_count, pitch, 2, projection, &seg_list);
word_box = blob_it->data()->bounding_box();
for (; blob_count > 0; blob_count--)
word_box += box_next(blob_it);
seg_it.set_to_list(&seg_list);
for (seg_it.mark_cycle_pt(); !seg_it.cycled_list(); seg_it.forward()) {
segpt = seg_it.data();
if (segpt->faked) {
colour = ScrollView::WHITE;
win->Pen(colour);
}
else {
win->Pen(colour);
}
win->Line(segpt->position(), word_box.bottom(), segpt->position(), word_box.top());
}
}
double check_pitch_sync( //find segmentation
BLOBNBOX_IT *blob_it, //blobs to do
int16_t blob_count, //no of blobs
int16_t pitch, //pitch estimate
int16_t pitch_error, //tolerance
STATS *projection, //vertical
FPSEGPT_LIST *seg_list //output list
) {
int16_t x; //current coord
int16_t min_index; //blob number
int16_t max_index; //blob number
int16_t left_edge; //of word
int16_t right_edge; //of word
int16_t right_max; //max allowed x
int16_t min_x; //in this region
int16_t max_x;
int16_t region_index;
int16_t best_region_index = 0; //for best result
int16_t offset; //dist to legal area
int16_t left_best_x; //edge of good region
int16_t right_best_x; //right edge
TBOX min_box; //bounding box
TBOX max_box; //bounding box
TBOX next_box; //box of next blob
FPSEGPT *segpt; //segment point
FPSEGPT_LIST *segpts; //points in a segment
double best_cost; //best path
double mean_sum; //computes result
FPSEGPT *best_end; //end of best path
BLOBNBOX_IT min_it; //copy iterator
BLOBNBOX_IT max_it; //copy iterator
FPSEGPT_IT segpt_it; //iterator
//output segments
FPSEGPT_IT outseg_it = seg_list;
FPSEGPT_LIST_CLIST lattice; //list of lists
//region iterator
FPSEGPT_LIST_C_IT lattice_it = &lattice;
// tprintf("Computing sync on word of %d blobs with pitch %d\n",
// blob_count, pitch);
// if (blob_count==8 && pitch==27)
// projection->print(stdout,TRUE);
if (pitch < 3)
pitch = 3; //nothing ludicrous
if ((pitch - 3) / 2 < pitch_error)
pitch_error = (pitch - 3) / 2;
min_it = *blob_it;
min_box = box_next (&min_it); //get box
// if (blob_count==8 && pitch==27)
// tprintf("1st box at (%d,%d)->(%d,%d)\n",
// min_box.left(),min_box.bottom(),
// min_box.right(),min_box.top());
//left of word
left_edge = min_box.left () + pitch_error;
for (min_index = 1; min_index < blob_count; min_index++) {
min_box = box_next (&min_it);
// if (blob_count==8 && pitch==27)
// tprintf("Box at (%d,%d)->(%d,%d)\n",
// min_box.left(),min_box.bottom(),
// min_box.right(),min_box.top());
}
right_edge = min_box.right (); //end of word
max_x = left_edge;
//min permissible
min_x = max_x - pitch + pitch_error * 2 + 1;
right_max = right_edge + pitch - pitch_error - 1;
segpts = new FPSEGPT_LIST; //list of points
segpt_it.set_to_list (segpts);
for (x = min_x; x <= max_x; x++) {
segpt = new FPSEGPT (x); //make a new one
//put in list
segpt_it.add_after_then_move (segpt);
}
//first segment
lattice_it.add_before_then_move (segpts);
min_index = 0;
region_index = 1;
best_cost = FLT_MAX;
best_end = nullptr;
min_it = *blob_it;
min_box = box_next (&min_it); //first box
do {
left_best_x = -1;
right_best_x = -1;
segpts = new FPSEGPT_LIST; //list of points
segpt_it.set_to_list (segpts);
min_x += pitch - pitch_error;//next limits
max_x += pitch + pitch_error;
while (min_box.right () < min_x && min_index < blob_count) {
min_index++;
min_box = box_next (&min_it);
}
max_it = min_it;
max_index = min_index;
max_box = min_box;
next_box = box_next (&max_it);
for (x = min_x; x <= max_x && x <= right_max; x++) {
while (x < right_edge && max_index < blob_count
&& x > max_box.right ()) {
max_index++;
max_box = next_box;
next_box = box_next (&max_it);
}
if (x <= max_box.left () + pitch_error
|| x >= max_box.right () - pitch_error || x >= right_edge
|| (max_index < blob_count - 1 && x >= next_box.left ())
|| (x - max_box.left () > pitch * pitsync_joined_edge
&& max_box.right () - x > pitch * pitsync_joined_edge)) {
// || projection->local_min(x))
if (x - max_box.left () > 0
&& x - max_box.left () <= pitch_error)
//dist to real break
offset = x - max_box.left ();
else if (max_box.right () - x > 0
&& max_box.right () - x <= pitch_error
&& (max_index >= blob_count - 1
|| x < next_box.left ()))
offset = max_box.right () - x;
else
offset = 0;
// offset=pitsync_offset_freecut_fraction*projection->pile_count(x);
segpt = new FPSEGPT (x, FALSE, offset, region_index,
pitch, pitch_error, lattice_it.data ());
}
else {
offset = projection->pile_count (x);
segpt = new FPSEGPT (x, TRUE, offset, region_index,
pitch, pitch_error, lattice_it.data ());
}
if (segpt->previous () != nullptr) {
segpt_it.add_after_then_move (segpt);
if (x >= right_edge - pitch_error) {
segpt->terminal = TRUE;//no more wanted
if (segpt->cost_function () < best_cost) {
best_cost = segpt->cost_function ();
//find least
best_end = segpt;
best_region_index = region_index;
left_best_x = x;
right_best_x = x;
}
else if (segpt->cost_function () == best_cost
&& right_best_x == x - 1)
right_best_x = x;
}
}
else {
delete segpt; //no good
}
}
if (segpts->empty ()) {
if (best_end != nullptr)
break; //already found one
make_illegal_segment (lattice_it.data (), min_box, min_it,
region_index, pitch, pitch_error, segpts);
}
else {
if (right_best_x > left_best_x + 1) {
left_best_x = (left_best_x + right_best_x + 1) / 2;
for (segpt_it.mark_cycle_pt (); !segpt_it.cycled_list ()
&& segpt_it.data ()->position () != left_best_x;
segpt_it.forward ());
if (segpt_it.data ()->position () == left_best_x)
//middle of region
best_end = segpt_it.data ();
}
}
//new segment
lattice_it.add_before_then_move (segpts);
region_index++;
}
while (min_x < right_edge);
ASSERT_HOST (best_end != nullptr);//must always find some
for (lattice_it.mark_cycle_pt (); !lattice_it.cycled_list ();
lattice_it.forward ()) {
segpts = lattice_it.data ();
segpt_it.set_to_list (segpts);
// if (blob_count==8 && pitch==27)
// {
// for (segpt_it.mark_cycle_pt();!segpt_it.cycled_list();segpt_it.forward())
// {
// segpt=segpt_it.data();
// tprintf("At %d, (%x) cost=%g, m=%g, sq=%g, pred=%x\n",
// segpt->position(),segpt,segpt->cost_function(),
// segpt->sum(),segpt->squares(),segpt->previous());
// }
// tprintf("\n");
// }
for (segpt_it.mark_cycle_pt (); !segpt_it.cycled_list ()
&& segpt_it.data () != best_end; segpt_it.forward ());
if (segpt_it.data () == best_end) {
//save good one
segpt = segpt_it.extract ();
outseg_it.add_before_then_move (segpt);
best_end = segpt->previous ();
}
}
ASSERT_HOST (best_end == nullptr);
ASSERT_HOST (!outseg_it.empty ());
outseg_it.move_to_last ();
mean_sum = outseg_it.data ()->sum ();
mean_sum = mean_sum * mean_sum / best_region_index;
if (outseg_it.data ()->squares () - mean_sum < 0)
tprintf ("Impossible sqsum=%g, mean=%g, total=%d\n",
outseg_it.data ()->squares (), outseg_it.data ()->sum (),
best_region_index);
lattice.deep_clear (); //shift the lot
return outseg_it.data ()->squares () - mean_sum;
}
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");
}