|
Tesseract
3.02
|
|
Tesseract
3.02
|
#include <edgblob.h>
Public Member Functions | |
| ~OL_BUCKETS () | |
| C_OUTLINE_LIST * | start_scan () |
| C_OUTLINE_LIST * | scan_next () |
OL_BUCKETS::OL_BUCKETS | |
Construct an array of buckets for associating outlines into blobs. | |
| OL_BUCKETS (ICOORD bleft, ICOORD tright) | |
OL_BUCKETS::operator( | |
Return a pointer to a list of C_OUTLINEs corresponding to the given pixel coordinates. | |
| C_OUTLINE_LIST * | operator() (inT16 x, inT16 y) |
OL_BUCKETS::count_children | |
Find number of descendants of this outline. | |
| inT32 | count_children (C_OUTLINE *outline, inT32 max_count) |
OL_BUCKETS::outline_complexity | |
This is the new version of count_child. The goal of this function is to determine if an outline and its interiors could be part of a character blob. This is done by computing a "complexity" index for the outline, which is the return value of this function, and checking it against a threshold. The max_count is used for short-circuiting the recursion and forcing a rejection that guarantees to fail the threshold test. The complexity F for outline X with N children X[i] is F(X) = N + sum_i F(X[i]) * edges_children_per_grandchild so each layer of nesting increases complexity exponentially. An outline can be rejected as a text blob candidate if its complexity is too high, has too many children(likely a container), or has too many layers of nested inner loops. This has the side-effect of flattening out boxed or reversed video text regions. | |
| inT32 | outline_complexity (C_OUTLINE *outline, inT32 max_count, inT16 depth) |
OL_BUCKETS::extract_children | |
Find number of descendants of this outline. | |
| void | extract_children (C_OUTLINE *outline, C_OUTLINE_IT *it) |
Definition at line 69 of file edgblob.cpp.
: bl(bleft), tr(tright) {
bxdim =(tright.x() - bleft.x()) / BUCKETSIZE + 1;
bydim =(tright.y() - bleft.y()) / BUCKETSIZE + 1;
// make array
buckets = new C_OUTLINE_LIST[bxdim * bydim];
index = 0;
}
| OL_BUCKETS::~OL_BUCKETS | ( | ) | [inline] |
Definition at line 184 of file edgblob.cpp.
{
BOOL8 parent_box; // could it be boxy
inT16 xmin, xmax; // coord limits
inT16 ymin, ymax;
inT16 xindex, yindex; // current bucket
C_OUTLINE *child; // current child
inT32 child_count; // no of children
inT32 grandchild_count; // no of grandchildren
inT32 parent_area; // potential box
FLOAT32 max_parent_area; // potential box
inT32 child_area; // current child
inT32 child_length; // current child
TBOX olbox;
C_OUTLINE_IT child_it; // search iterator
olbox = outline->bounding_box();
xmin =(olbox.left() - bl.x()) / BUCKETSIZE;
xmax =(olbox.right() - bl.x()) / BUCKETSIZE;
ymin =(olbox.bottom() - bl.y()) / BUCKETSIZE;
ymax =(olbox.top() - bl.y()) / BUCKETSIZE;
child_count = 0;
grandchild_count = 0;
parent_area = 0;
max_parent_area = 0;
parent_box = TRUE;
for (yindex = ymin; yindex <= ymax; yindex++) {
for (xindex = xmin; xindex <= xmax; xindex++) {
child_it.set_to_list(&buckets[yindex * bxdim + xindex]);
if (child_it.empty())
continue;
for (child_it.mark_cycle_pt(); !child_it.cycled_list();
child_it.forward()) {
child = child_it.data();
if (child != outline && *child < *outline) {
child_count++;
if (child_count <= max_count) {
int max_grand =(max_count - child_count) /
edges_children_per_grandchild;
if (max_grand > 0)
grandchild_count += count_children(child, max_grand) *
edges_children_per_grandchild;
else
grandchild_count += count_children(child, 1);
}
if (child_count + grandchild_count > max_count) {
if (edges_debug)
tprintf("Discarding parent with child count=%d, gc=%d\n",
child_count,grandchild_count);
return child_count + grandchild_count;
}
if (parent_area == 0) {
parent_area = outline->outer_area();
if (parent_area < 0)
parent_area = -parent_area;
max_parent_area = outline->bounding_box().area() * edges_boxarea;
if (parent_area < max_parent_area)
parent_box = FALSE;
}
if (parent_box &&
(!edges_children_fix ||
child->bounding_box().height() > edges_min_nonhole)) {
child_area = child->outer_area();
if (child_area < 0)
child_area = -child_area;
if (edges_children_fix) {
if (parent_area - child_area < max_parent_area) {
parent_box = FALSE;
continue;
}
if (grandchild_count > 0) {
if (edges_debug)
tprintf("Discarding parent of area %d, child area=%d, max%g "
"with gc=%d\n",
parent_area, child_area, max_parent_area,
grandchild_count);
return max_count + 1;
}
child_length = child->pathlength();
if (child_length * child_length >
child_area * edges_patharea_ratio) {
if (edges_debug)
tprintf("Discarding parent of area %d, child area=%d, max%g "
"with child length=%d\n",
parent_area, child_area, max_parent_area,
child_length);
return max_count + 1;
}
}
if (child_area < child->bounding_box().area() * edges_childarea) {
if (edges_debug)
tprintf("Discarding parent of area %d, child area=%d, max%g "
"with child rect=%d\n",
parent_area, child_area, max_parent_area,
child->bounding_box().area());
return max_count + 1;
}
}
}
}
}
}
return child_count + grandchild_count;
}
| void OL_BUCKETS::extract_children | ( | C_OUTLINE * | outline, |
| C_OUTLINE_IT * | it | ||
| ) |
Definition at line 300 of file edgblob.cpp.
{
inT16 xmin, xmax; // coord limits
inT16 ymin, ymax;
inT16 xindex, yindex; // current bucket
TBOX olbox;
C_OUTLINE_IT child_it; // search iterator
olbox = outline->bounding_box();
xmin =(olbox.left() - bl.x()) / BUCKETSIZE;
xmax =(olbox.right() - bl.x()) / BUCKETSIZE;
ymin =(olbox.bottom() - bl.y()) / BUCKETSIZE;
ymax =(olbox.top() - bl.y()) / BUCKETSIZE;
for (yindex = ymin; yindex <= ymax; yindex++) {
for (xindex = xmin; xindex <= xmax; xindex++) {
child_it.set_to_list(&buckets[yindex * bxdim + xindex]);
for (child_it.mark_cycle_pt(); !child_it.cycled_list();
child_it.forward()) {
if (*child_it.data() < *outline) {
it->add_after_then_move(child_it.extract());
}
}
}
}
}
Definition at line 88 of file edgblob.cpp.
{
return &buckets[(y-bl.y()) / BUCKETSIZE * bxdim + (x-bl.x()) / BUCKETSIZE];
}
Definition at line 115 of file edgblob.cpp.
{
inT16 xmin, xmax; // coord limits
inT16 ymin, ymax;
inT16 xindex, yindex; // current bucket
C_OUTLINE *child; // current child
inT32 child_count; // no of children
inT32 grandchild_count; // no of grandchildren
C_OUTLINE_IT child_it; // search iterator
TBOX olbox = outline->bounding_box();
xmin =(olbox.left() - bl.x()) / BUCKETSIZE;
xmax =(olbox.right() - bl.x()) / BUCKETSIZE;
ymin =(olbox.bottom() - bl.y()) / BUCKETSIZE;
ymax =(olbox.top() - bl.y()) / BUCKETSIZE;
child_count = 0;
grandchild_count = 0;
if (++depth > edges_max_children_layers) // nested loops are too deep
return max_count + depth;
for (yindex = ymin; yindex <= ymax; yindex++) {
for (xindex = xmin; xindex <= xmax; xindex++) {
child_it.set_to_list(&buckets[yindex * bxdim + xindex]);
if (child_it.empty())
continue;
for (child_it.mark_cycle_pt(); !child_it.cycled_list();
child_it.forward()) {
child = child_it.data();
if (child == outline || !(*child < *outline))
continue;
child_count++;
if (child_count > edges_max_children_per_outline) { // too fragmented
if (edges_debug)
tprintf("Discard outline on child_count=%d > "
"max_children_per_outline=%d\n",
child_count,
static_cast<inT32>(edges_max_children_per_outline));
return max_count + child_count;
}
// Compute the "complexity" of each child recursively
inT32 remaining_count = max_count - child_count - grandchild_count;
if (remaining_count > 0)
grandchild_count += edges_children_per_grandchild *
outline_complexity(child, remaining_count, depth);
if (child_count + grandchild_count > max_count) { // too complex
if (edges_debug)
tprintf("Disgard outline on child_count=%d + grandchild_count=%d "
"> max_count=%d\n",
child_count, grandchild_count, max_count);
return child_count + grandchild_count;
}
}
}
}
return child_count + grandchild_count;
}
| C_OUTLINE_LIST* OL_BUCKETS::scan_next | ( | ) | [inline] |
| C_OUTLINE_LIST* OL_BUCKETS::start_scan | ( | ) | [inline] |
1.7.5.1