Olena-patches June 2013

olena-patches@lrde.epita.fr

5 participants
104 discussions

olena: olena-2.0-571-g929c08c Create blured copies of original image

by Anthony Seure

--- milena/sandbox/anthony/Makefile | 11 ++- milena/sandbox/anthony/src/scale_space.cc | 114 ++++++++++++++++++++++++----- 2 files changed, 103 insertions(+), 22 deletions(-) diff --git a/milena/sandbox/anthony/Makefile b/milena/sandbox/anthony/Makefile index 7f990da..81f788e 100644 --- a/milena/sandbox/anthony/Makefile +++ b/milena/sandbox/anthony/Makefile @@ -1,8 +1,8 @@ CCACHE=ccache CC=g++ -CFLAGS=-Wall -Werror -O3 -DNDEBUG -CLIBS=-I../../../scribo/ -I../../ -CLEAN=*.o output/* log final.xml *.pgm +CFLAGS=-Wall -Werror +CLIBS=-I../../ +CLEAN=*.o output/* log SRC=src/scale_space.cc OUTPUT=a.out @@ -10,7 +10,10 @@ OUTPUT=a.out all: scale scale: - $(CCACHE) $(CC) $(CFLAGS) $(CLIBS) $(SRC) -o $(OUTPUT) + $(CCACHE) $(CC) $(CFLAGS) -O2 -DNDEBUG $(CLIBS) $(SRC) -o $(OUTPUT) + +debug: + $(CCACHE) $(CC) $(CFLAGS) -g -ggdb $(CLIBS) $(SRC) -o $(OUTPUT) clean: rm -rf $(CLEAN) diff --git a/milena/sandbox/anthony/src/scale_space.cc b/milena/sandbox/anthony/src/scale_space.cc index 9e452a3..31180af 100644 --- a/milena/sandbox/anthony/src/scale_space.cc +++ b/milena/sandbox/anthony/src/scale_space.cc @@ -1,36 +1,114 @@ -#include <mln/binarization/all.hh> - #include <mln/core/image/image2d.hh> - +#include <mln/core/alias/neighb2d.hh> #include <mln/data/all.hh> -#include <mln/draw/line.hh> +#include <mln/io/pgm/all.hh> +#include <mln/literal/all.hh> +#include <mln/value/all.hh> +#include <mln/make/w_window2d.hh> +#include <mln/core/alias/w_window2d_int.hh> -#include <mln/fun/v2v/rgb_to_luma.hh> +#include <cmath> -#include <mln/io/pgm/all.hh> +using namespace mln; -#include <mln/labeling/all.hh> -#include <mln/literal/all.hh> -#include <mln/logical/and.hh> +// Set the gaussian kernel of size (6 * t + 1) +double *gaussian_kernel(unsigned t) +{ + int size = 6 * t + 1; + unsigned index = 0; + double *kernel = new double[size * size]; + static const double pi = 4.0 * atan(1.0); + double div = (2.0 * pi * t * t); -#include <mln/value/all.hh> + for (int i = -(size / 2); i < (size / 2) + 1; ++i) + { + for (int j = -(size / 2); j < (size / 2) + 1; ++j) + { + double e = exp(- (i * i + j * j) / (2.0 * t * t) ); + kernel[index++] = e / div; + } + } -#include <tesseract/baseapi.h> + return kernel; +} -#include <scribo/binarization/sauvola.hh> -#include <scribo/core/component_set.hh> -#include <scribo/preprocessing/denoise_fg.hh> -#include <scribo/primitive/extract/all.hh> +// Apply the convolution of the image by the kernel +template<typename I> +void convolve(const I& original, I& filtered, double *kernel, unsigned t) +{ + mln_piter(I) p(filtered.domain()); + int size = 6 * t + 1; + window2d win; -using namespace mln; + for (int i = -(size / 2); i < (size / 2) + 1; ++i) + { + for (int j = -(size / 2); j < (size / 2) + 1; ++j) + { + win.insert(i, j); + } + } + + + // Iterate through all image sites + for_all(p) + { + // Create the window around the site + mln_qiter(window2d) q(win, p); + float sum = 0; + int index = 0; + // Iterate through all window image sites + for_all(q) + { + if (filtered.has(q)) + sum += original(q) * kernel[index++]; + else + sum += 127 * kernel[index++]; + } + + filtered(p) = static_cast<int>(sum); + } +} + +// Blur the image thanks to a convolution matrix +template<typename I> +image2d<value::int_u8> blur(const I& original, unsigned t) +{ + I filtered; + double *kernel = gaussian_kernel(t); + + initialize(filtered, original); + convolve(original, filtered, kernel, t); + + return filtered; +} + +template<typename I> +void downscaleResolution(const I& original, I& toReduce) +{ +} + +// MAIN ENTRY POINT int main(int argc, char** argv) { image2d<value::int_u8> original; + io::pgm::load(original, "images/flower.pgm"); + + image2d<value::int_u8> blur1, blur2, blur4; + initialize(blur1, original); + initialize(blur2, original); + initialize(blur4, original); + + io::pgm::save(original, "output/original.pgm"); + + blur1 = blur(original, 1); + io::pgm::save(blur1, "output/blur1.pgm"); - io::pgm::load(original, "images/keith.pgm"); + blur2 = blur(original, 2); + io::pgm::save(blur2, "output/blur2.pgm"); - io::pgm::save(original, "final.pgm"); + blur4 = blur(original, 4); + io::pgm::save(blur4, "output/blur4.pgm"); return 0; } -- 1.7.2.5

10 years, 11 months

olena: olena-2.0-570-g6836fdc Init scale space working directory

by Anthony Seure

--- milena/sandbox/anthony/Makefile | 21 +++++++++++++++++ milena/sandbox/anthony/src/scale_space.cc | 36 +++++++++++++++++++++++++++++ 2 files changed, 57 insertions(+), 0 deletions(-) create mode 100644 milena/sandbox/anthony/Makefile create mode 100644 milena/sandbox/anthony/src/scale_space.cc diff --git a/milena/sandbox/anthony/Makefile b/milena/sandbox/anthony/Makefile new file mode 100644 index 0000000..7f990da --- /dev/null +++ b/milena/sandbox/anthony/Makefile @@ -0,0 +1,21 @@ +CCACHE=ccache +CC=g++ +CFLAGS=-Wall -Werror -O3 -DNDEBUG +CLIBS=-I../../../scribo/ -I../../ +CLEAN=*.o output/* log final.xml *.pgm + +SRC=src/scale_space.cc +OUTPUT=a.out + +all: scale + +scale: + $(CCACHE) $(CC) $(CFLAGS) $(CLIBS) $(SRC) -o $(OUTPUT) + +clean: + rm -rf $(CLEAN) + +mrproper: clean + rm -f $(OUTPUT) + +.PHONY: scale clean mrproper diff --git a/milena/sandbox/anthony/src/scale_space.cc b/milena/sandbox/anthony/src/scale_space.cc new file mode 100644 index 0000000..9e452a3 --- /dev/null +++ b/milena/sandbox/anthony/src/scale_space.cc @@ -0,0 +1,36 @@ +#include <mln/binarization/all.hh> + +#include <mln/core/image/image2d.hh> + +#include <mln/data/all.hh> +#include <mln/draw/line.hh> + +#include <mln/fun/v2v/rgb_to_luma.hh> + +#include <mln/io/pgm/all.hh> + +#include <mln/labeling/all.hh> +#include <mln/literal/all.hh> +#include <mln/logical/and.hh> + +#include <mln/value/all.hh> + +#include <tesseract/baseapi.h> + +#include <scribo/binarization/sauvola.hh> +#include <scribo/core/component_set.hh> +#include <scribo/preprocessing/denoise_fg.hh> +#include <scribo/primitive/extract/all.hh> + +using namespace mln; + +int main(int argc, char** argv) +{ + image2d<value::int_u8> original; + + io::pgm::load(original, "images/keith.pgm"); + + io::pgm::save(original, "final.pgm"); + + return 0; +} -- 1.7.2.5

10 years, 11 months

olena: olena-2.0-569-g4ecdb6a Remove many false positives (intersecting boxes)

by Anthony Seure

--- scribo/sandbox/icdar_13_table/src/new.cc | 35 +++++++++++++++++++++++++++++- 1 files changed, 34 insertions(+), 1 deletions(-) diff --git a/scribo/sandbox/icdar_13_table/src/new.cc b/scribo/sandbox/icdar_13_table/src/new.cc index bcf127e..d52ac15 100644 --- a/scribo/sandbox/icdar_13_table/src/new.cc +++ b/scribo/sandbox/icdar_13_table/src/new.cc @@ -58,6 +58,13 @@ using namespace mln; +// Check if two boxes intersected +bool intersect(const box2d& b1, const box2d& b2) +{ + return ((unsigned) abs(b1.pcenter().col() - b2.pcenter().col()) * 2 < b1.width() + b2.width() + && (unsigned) abs(b1.pcenter().row() - b2.pcenter().row()) * 2 < b1.height() + b2.height()); +} + // Draw weighted boxes (red < orange < cyan < green) // 1 link < 2 links < 3 links < 3+ links template<typename T, typename L> @@ -581,6 +588,9 @@ int main(int argc, char** argv) descendants[columns_set.find(j)].insert(j); } + std::vector<box2d> columns; + std::set<unsigned> not_valid; + // Visualization: columns. image2d<value::rgb8> ima_columns = data::convert(value::rgb8(), bin_merged); for (unsigned i = 0; i < groups.nelements(); ++i) @@ -590,9 +600,32 @@ int main(int argc, char** argv) for (group_set::const_iterator j = descendants[i].begin(); j != descendants[i].end(); ++j) column_box.merge(groups(*j).bbox()); - draw::box(ima_columns, column_box, literal::red); + draw::box(ima_columns, column_box, literal::green); + columns.push_back(column_box); } + for (unsigned i = 0; i < columns.size(); ++i) + { + for (unsigned j = 0; j < columns.size(); ++j) + { + if (i != j) + { + const box2d& c1 = columns[i]; + const box2d& c2 = columns[j]; + + if (intersect(c1, c2) || intersect(c2, c1)) + { + not_valid.insert(i); + not_valid.insert(j); + } + } + } + } + + for (unsigned i = 0; i < columns.size(); ++i) + if (not_valid.find(i) != not_valid.end()) + draw::box(ima_columns, columns[i], literal::red); + // Write images and close XML std::ostringstream path; unsigned number = 0; -- 1.7.2.5

10 years, 11 months

olena: olena-2.0-568-gc9c37e9 Build the set of detected columns during table recognition.

by Roland Levillain

* scribo/sandbox/icdar_13_table/src/new.cc: Here. * scribo/sandbox/icdar_13_table/src/disjoint_set.hh: New. --- scribo/sandbox/icdar_13_table/src/disjoint_set.hh | 54 +++++++++++++++++++++ scribo/sandbox/icdar_13_table/src/new.cc | 36 ++++++++++++++ 2 files changed, 90 insertions(+), 0 deletions(-) create mode 100644 scribo/sandbox/icdar_13_table/src/disjoint_set.hh diff --git a/scribo/sandbox/icdar_13_table/src/disjoint_set.hh b/scribo/sandbox/icdar_13_table/src/disjoint_set.hh new file mode 100644 index 0000000..b8ae409 --- /dev/null +++ b/scribo/sandbox/icdar_13_table/src/disjoint_set.hh @@ -0,0 +1,54 @@ +#ifndef DISJOINT_SET_HH +# define DISJOINT_SET_HH + +/// \file +/// Implementation of a disjoint-set forest. + +# include <limits> +# include <vector> + +template <typename T> +struct disjoint_set +{ + disjoint_set(size_t size) + // The maximum value of T is considered an ``invalid value''. + : parent(size, std::numeric_limits<T>::max()) + { + for (unsigned i = 0; i < parent.size(); ++i) + make_set(i); + } + + /// Create initial sets (singletons). + void make_set(T x) + { + parent[x] = x; + } + + bool is_root(T x) + { + return parent[x] == x; + } + + /// Find the root of the set containing \a x. + T find(T x) + { + if (!is_root(x)) + // Path compression. + parent[x] = find(parent[x]); + return parent[x]; + } + + /// Merge the set containing \a y into the set containing \a x. + void make_union(T x, T y) + { + T x_root = find(x); + T y_root = find(y); + parent[y_root] = x_root; + } + + /// Parent relationship. A node with is its own parent is a + /// root (representative of a component). + std::vector<T> parent; +}; + +#endif // ! DISJOINT_SET_HH diff --git a/scribo/sandbox/icdar_13_table/src/new.cc b/scribo/sandbox/icdar_13_table/src/new.cc index 2c41e31..bcf127e 100644 --- a/scribo/sandbox/icdar_13_table/src/new.cc +++ b/scribo/sandbox/icdar_13_table/src/new.cc @@ -54,6 +54,8 @@ #include <scribo/text/extract_paragraphs.hh> #include <scribo/text/merging.hh> +#include "disjoint_set.hh" + using namespace mln; // Draw weighted boxes (red < orange < cyan < green) @@ -558,6 +560,39 @@ int main(int argc, char** argv) } } + // Build the columns. + // First pass. + disjoint_set<node_id> columns_set(groups.nelements()); + for (unsigned i = 0; i < groups.nelements(); ++i) + { + const group_set& successors = nodes_below[i]; + for (group_set::const_iterator j = successors.begin(); + j != successors.end(); ++j) + columns_set.make_union(i, *j); + } + // Second pass: assign labels. Label 0 is unused and means + // ``default''. + std::vector<group_set> descendants(groups.nelements()); + for (unsigned i = 0; i < groups.nelements(); ++i) + { + // Process groups in reverse order. + unsigned j = groups.nelements() - 1 - i; + if (!columns_set.is_root(j)) + descendants[columns_set.find(j)].insert(j); + } + + // Visualization: columns. + image2d<value::rgb8> ima_columns = data::convert(value::rgb8(), bin_merged); + for (unsigned i = 0; i < groups.nelements(); ++i) + if (!descendants[i].empty()) + { + box2d column_box = groups(i).bbox(); + for (group_set::const_iterator j = descendants[i].begin(); + j != descendants[i].end(); ++j) + column_box.merge(groups(*j).bbox()); + draw::box(ima_columns, column_box, literal::red); + } + // Write images and close XML std::ostringstream path; unsigned number = 0; @@ -572,6 +607,7 @@ int main(int argc, char** argv) write_image(ima_links, "components", page, number, path); write_image(ima_groups, "groups", page, number, path); write_image(ima_valid, "valid", page, number, path); + write_image(ima_columns, "columns", page, number, path); } delete xml; -- 1.7.2.5

10 years, 11 months

olena: olena-2.0-567-g5ead1be Build spatial relations between text boxes as directed graphs.

by Roland Levillain

* scribo/sandbox/icdar_13_table/src/new.cc: Here. --- scribo/sandbox/icdar_13_table/src/new.cc | 89 ++++++++++++++++++++++++++++++ 1 files changed, 89 insertions(+), 0 deletions(-) diff --git a/scribo/sandbox/icdar_13_table/src/new.cc b/scribo/sandbox/icdar_13_table/src/new.cc index 02e7594..2c41e31 100644 --- a/scribo/sandbox/icdar_13_table/src/new.cc +++ b/scribo/sandbox/icdar_13_table/src/new.cc @@ -469,6 +469,95 @@ int main(int argc, char** argv) draw_links_rl(groups, ima_groups, balance, vlines); draw_adjacency_boxes(balance, ima_valid, groups); + /* FIXME: The code below duplicates some of the code in the + draw_links_* routines. Factor. */ + // Adjacencies between nodes. Of course, an actual digraph data + // structure would be better. + typedef unsigned node_id; + typedef std::set<node_id> group_set; + typedef std::vector<group_set> adjacencies; + adjacencies nodes_below(groups.nelements()); + adjacencies nodes_above(groups.nelements()); + adjacencies nodes_right(groups.nelements()); + adjacencies nodes_left(groups.nelements()); + + // Draw vertical links (red) + for (unsigned i = 1; i < groups.nelements(); ++i) + { + for (unsigned j = i + 1; j < groups.nelements(); ++j) + { + const box2d& b1 = groups(i).bbox(); + const box2d& b2 = groups(j).bbox(); + const point2d& p1 = b1.pcenter(); + const point2d& p2 = b2.pcenter(); + + unsigned max_height = std::max(b1.height(), b2.height()); + unsigned min_height = std::min(b1.height(), b2.height()); + + if (/* p1[0] < p2[0] // Avoid redundancy + && */ + max_height * 2 < bin_merged.ncols() + && min_height + 3 >= max_height // Same heights + && b1.width() < 2 * average_width && b2.width() < 2 * average_width // Regular width + && (b1.pmin()[1] == b2.pmin()[1] + || (b1.pmin()[1] < b2.pmin()[1] && b1.pmax()[1] > b2.pmin()[1]) + || (b1.pmin()[1] > b2.pmin()[1] && b2.pmax()[1] > b1.pmin()[1])) // Boxes are aligned + && abs(p1[0] - p2[0]) < 3 * max_height // Reduced gap + && abs(p1[1] - p2[1]) < 20) // Vertical proximity + { + // Build the above/below adjacencies. + node_id top_node, bottom_node; + if (p1.row() < p2.col()) + { + top_node = i; + bottom_node = j; + } + else + { + top_node = j; + bottom_node = i; + } + nodes_below[top_node].insert(bottom_node); + nodes_above[bottom_node].insert(top_node); + } + } + } + + // Draw horizontal links (green) + for (unsigned i = 1; i < groups.nelements(); ++i) + { + for (unsigned j = i + 1; j < groups.nelements(); ++j) + { + const box2d& b1 = groups(i).bbox(); + const box2d& b2 = groups(j).bbox(); + const point2d& p1 = b1.pcenter(); + const point2d& p2 = b2.pcenter(); + + if (/* p1[1] < p2[1] // Avoid redundancy + && */ + (b1.pmin()[0] == b2.pmin()[0] + || (b1.pmin()[0] < b2.pmin()[0] && b1.pmax()[0] > b2.pmin()[0]) + || (b1.pmin()[0] > b2.pmin()[0] && b2.pmax()[0] > b1.pmin()[0])) // Boxes are aligned + && abs(p1[0] - p2[0]) < 10) // Reduced gap + { + // Build the right/left adjacencies. + node_id left_node, right_node; + if (p1.col() < p2.col()) + { + left_node = i; + right_node = j; + } + else + { + left_node = j; + right_node = i; + } + nodes_right[left_node].insert(right_node); + nodes_left[right_node].insert(left_node); + } + } + } + // Write images and close XML std::ostringstream path; unsigned number = 0; -- 1.7.2.5

10 years, 11 months

olena: olena-2.0-566-g6117c48 Fix various issues in the table detection experiment.

by Roland Levillain

* scribo/sandbox/icdar_13_table/src/new.cc: Use the multiscale version of the Sauvola binarization. (main): Remove an unused variable. Remove superfluous `return' statement. --- scribo/sandbox/icdar_13_table/src/new.cc | 7 +------ 1 files changed, 1 insertions(+), 6 deletions(-) diff --git a/scribo/sandbox/icdar_13_table/src/new.cc b/scribo/sandbox/icdar_13_table/src/new.cc index b13696e..02e7594 100644 --- a/scribo/sandbox/icdar_13_table/src/new.cc +++ b/scribo/sandbox/icdar_13_table/src/new.cc @@ -26,7 +26,6 @@ #include <tesseract/baseapi.h> // INCLUDES SCRIBO -#include <scribo/binarization/sauvola.hh> #include <scribo/binarization/sauvola_ms.hh> #include <scribo/core/component_set.hh> @@ -353,8 +352,7 @@ int main(int argc, char** argv) { I8 original = pdf[page]; image2d<value::int_u8> filtered = data::transform(original, fun::v2v::rgb_to_luma<value::int_u8>()); - IB bin = scribo::binarization::sauvola(filtered, 81, 0.44); - //IB bin = scribo::binarization::sauvola_ms(filtered, 81, 2); + IB bin = scribo::binarization::sauvola_ms(filtered, 81, 2); // Reverse selection IB reverse = logical::not_(bin); @@ -378,7 +376,6 @@ int main(int argc, char** argv) // Find separators IB bin_without_separators = duplicate(bin); - IB separators = separators; V nhlines, nvlines; unsigned min_width = 31; unsigned min_height = 71; @@ -489,6 +486,4 @@ int main(int argc, char** argv) } delete xml; - - return 0; } -- 1.7.2.5

10 years, 11 months

olena: olena-2.0-565-g103edf9 [ICDAR_13] Refactore and use of sauvola_ms instead of sauvola

by Anthony Seure

--- scribo/sandbox/icdar_13_table/src/new.cc | 93 +++++++++++++++++------------- 1 files changed, 53 insertions(+), 40 deletions(-) diff --git a/scribo/sandbox/icdar_13_table/src/new.cc b/scribo/sandbox/icdar_13_table/src/new.cc index 5370164..b13696e 100644 --- a/scribo/sandbox/icdar_13_table/src/new.cc +++ b/scribo/sandbox/icdar_13_table/src/new.cc @@ -27,6 +27,7 @@ // INCLUDES SCRIBO #include <scribo/binarization/sauvola.hh> +#include <scribo/binarization/sauvola_ms.hh> #include <scribo/core/component_set.hh> #include <scribo/core/line_set.hh> @@ -56,6 +57,29 @@ using namespace mln; +// Draw weighted boxes (red < orange < cyan < green) +// 1 link < 2 links < 3 links < 3+ links +template<typename T, typename L> +void draw_adjacency_boxes(const std::vector<short>& balance, + image2d<T>& ima, + const scribo::object_groups< image2d<L> >& groups) +{ + for (unsigned i = 0; i < balance.size(); ++i) + { + if (balance[i] == 1) + draw::box(ima, groups(i).bbox(), literal::red); + + if (balance[i] == 2) + draw::box(ima, groups(i).bbox(), literal::orange); + + if (balance[i] == 3) + draw::box(ima, groups(i).bbox(), literal::cyan); + + if (balance[i] > 3) + draw::box(ima, groups(i).bbox(), literal::green); + } +} + // Write image2d<bool> images void write_image(const image2d<bool>& ima, const char* name, @@ -226,7 +250,8 @@ void draw_links_lr(const scribo::object_groups< image2d<unsigned> >& groups, && (b1.pmin()[0] == b2.pmin()[0] || (b1.pmin()[0] < b2.pmin()[0] && b1.pmax()[0] > b2.pmin()[0]) || (b1.pmin()[0] > b2.pmin()[0] && b2.pmax()[0] > b1.pmin()[0])) // Boxes are aligned - && abs(p1[0] - p2[0]) < 10) // Reduced gap + && abs(p1[0] - p2[0]) < 10 // Reduced gap + && abs(p1[1] - p2[1]) > (b1.width() + b2.width()) / 4) // Consistent gap { unsigned k = 1; short separators = 0; @@ -277,7 +302,8 @@ void draw_links_rl(const scribo::object_groups< image2d<unsigned> >& groups, && (b1.pmin()[0] == b2.pmin()[0] || (b1.pmin()[0] < b2.pmin()[0] && b1.pmax()[0] > b2.pmin()[0]) || (b1.pmin()[0] > b2.pmin()[0] && b2.pmax()[0] > b1.pmin()[0])) // Boxes are aligned - && abs(p1[0] - p2[0]) < 10) // Reduced gap + && abs(p1[0] - p2[0]) < 10 // Reduced gap + && abs(p1[1] - p2[1]) > (b1.width() + b2.width()) / 4) // Consistent gap { unsigned k = 1; short separators = 0; @@ -311,34 +337,33 @@ int main(int argc, char** argv) { typedef value::label_16 V; typedef image2d<V> L; - - std::ostringstream path; - image2d<value::rgb8> original, ima_links, ima_groups, ima_valid; - image2d<value::int_u8> filtered; - image2d<bool> bin, reverse, reverse_selection, bin_merged, separators, bin_without_separators, whitespaces, comp, denoised; - scribo::component_set< image2d<unsigned> > components, rcomponents; - - unsigned dpi = 72; + typedef image2d<value::rgb8> I8; + typedef image2d<bool> IB; + typedef scribo::component_set< image2d<unsigned> > CS; // Loading and binarization XML* xml = new XML("final.xml", argv[1]); util::array< image2d<value::rgb8> > pdf; + unsigned dpi = 72; io::pdf::load(pdf, argv[1], dpi); + // Iterate over all pages for (unsigned page = 0; page < pdf.nelements(); ++page) { - original = pdf[page]; - filtered = data::transform(original, fun::v2v::rgb_to_luma<value::int_u8>()); - bin = scribo::binarization::sauvola(filtered, 81, 0.44); + I8 original = pdf[page]; + image2d<value::int_u8> filtered = data::transform(original, fun::v2v::rgb_to_luma<value::int_u8>()); + IB bin = scribo::binarization::sauvola(filtered, 81, 0.44); + //IB bin = scribo::binarization::sauvola_ms(filtered, 81, 2); // Reverse selection - reverse = logical::not_(bin); + IB reverse = logical::not_(bin); + IB reverse_selection; initialize(reverse_selection, reverse); data::fill(reverse_selection, false); unsigned nrcomponents; - rcomponents = scribo::primitive::extract::components(reverse, c8(), nrcomponents); + CS rcomponents = scribo::primitive::extract::components(reverse, c8(), nrcomponents); for (unsigned i = 1; i < rcomponents.nelements(); ++i) { @@ -352,8 +377,8 @@ int main(int argc, char** argv) reverse_selection = scribo::preprocessing::denoise_fg(reverse_selection, c8(), 4); // Find separators - bin_without_separators = duplicate(bin); - separators = separators; + IB bin_without_separators = duplicate(bin); + IB separators = separators; V nhlines, nvlines; unsigned min_width = 31; unsigned min_height = 71; @@ -366,14 +391,15 @@ int main(int argc, char** argv) data::fill((bin_without_separators | vlines(i).bbox()).rw(), false); // Denoise - denoised = scribo::preprocessing::denoise_fg(bin_without_separators, c8(), 4); + IB denoised = scribo::preprocessing::denoise_fg(bin_without_separators, c8(), 4); // Bin merged - bin_merged = logical::or_(denoised, reverse_selection); + IB bin_merged = logical::or_(denoised, reverse_selection); // Extract components unsigned ncomponents; - components = scribo::primitive::extract::components(bin_merged, c8(), ncomponents); + CS components = scribo::primitive::extract::components(bin_merged, c8(), ncomponents); + IB comp; initialize(comp, bin_merged); data::fill(comp, false); @@ -398,9 +424,11 @@ int main(int argc, char** argv) // Filter links scribo::object_links< image2d<unsigned> > hratio_filtered_links = scribo::filter::object_links_bbox_h_ratio(merged_links, 2.5f); - ima_links = data::convert(value::rgb8(), bin_merged); - ima_groups = data::convert(value::rgb8(), bin_merged); - ima_valid = data::convert(value::rgb8(), bin_merged); + IB tmp = logical::and_(bin_merged, comp); + + I8 ima_links = data::convert(value::rgb8(), tmp); + I8 ima_groups = data::convert(value::rgb8(), tmp); + I8 ima_valid = data::convert(value::rgb8(), tmp); // Write links for (unsigned l = 1; l < merged_links.nelements(); ++l) @@ -442,25 +470,10 @@ int main(int argc, char** argv) draw_links_bt(groups, ima_groups, balance, average_width, hlines); draw_links_lr(groups, ima_groups, balance, vlines); draw_links_rl(groups, ima_groups, balance, vlines); - - // Draw weighted boxes (red < orange < cyan < green) - // 1 link < 2 links < 3 links < 3+ links - for (unsigned i = 0; i < balance.size(); ++i) - { - if (balance[i] == 1) - draw::box(ima_valid, groups(i).bbox(), literal::red); - - if (balance[i] == 2) - draw::box(ima_valid, groups(i).bbox(), literal::orange); - - if (balance[i] == 3) - draw::box(ima_valid, groups(i).bbox(), literal::cyan); - - if (balance[i] > 3) - draw::box(ima_valid, groups(i).bbox(), literal::green); - } + draw_adjacency_boxes(balance, ima_valid, groups); // Write images and close XML + std::ostringstream path; unsigned number = 0; write_image(bin, "bin", page, number, path); -- 1.7.2.5

10 years, 11 months

olena: olena-2.0-564-g3e47e4a [ICDAR_13] Validate links of boxes based on separators

by Anthony Seure

--- scribo/sandbox/icdar_13_table/src/new.cc | 123 +++++++++++++++++++++++++----- 1 files changed, 102 insertions(+), 21 deletions(-) diff --git a/scribo/sandbox/icdar_13_table/src/new.cc b/scribo/sandbox/icdar_13_table/src/new.cc index 9f05030..5370164 100644 --- a/scribo/sandbox/icdar_13_table/src/new.cc +++ b/scribo/sandbox/icdar_13_table/src/new.cc @@ -87,10 +87,12 @@ void write_image(const image2d<value::rgb8>& ima, } // Draw vertical links from top to bottom (red) +template<typename L> void draw_links_tb(const scribo::object_groups< image2d<unsigned> >& groups, image2d<value::rgb8>& ima_groups, std::vector<short>& balance, - unsigned average_width) + unsigned average_width, + const scribo::component_set<L>& hlines) { for (unsigned i = 1; i <= groups.nelements(); ++i) { @@ -116,9 +118,27 @@ void draw_links_tb(const scribo::object_groups< image2d<unsigned> >& groups, && (unsigned) abs(p1[0] - p2[0]) < 3 * max_height // Reduced gap && (unsigned) abs(p1[1] - p2[1]) < 20) // Vertical proximity { - draw::line(ima_groups, p1, p2, literal::red); - balance[i] += 1; - break; + unsigned k = 1; + short separators = 0; + + while (k <= hlines.nelements() && separators < 2) + { + const box2d& s = hlines(k).bbox(); + + if (s.pmin()[1] <= b1.pmin()[1] && s.pmax()[1] >= b1.pmax()[1] + && s.pmin()[0] > b1.pmax()[0] + && s.pmax()[0] < b2.pmin()[0]) + ++separators; + + ++k; + } + + if (separators < 2) + { + draw::line(ima_groups, p1, p2, literal::red); + balance[i] += 1; + break; + } } } } @@ -126,10 +146,12 @@ void draw_links_tb(const scribo::object_groups< image2d<unsigned> >& groups, } // Draw vertical links from bottom to top (red) +template<typename L> void draw_links_bt(const scribo::object_groups< image2d<unsigned> >& groups, image2d<value::rgb8>& ima_groups, std::vector<short>& balance, - unsigned average_width) + unsigned average_width, + const scribo::component_set<L>& hlines) { for (unsigned i = groups.nelements(); i > 0; --i) { @@ -155,9 +177,27 @@ void draw_links_bt(const scribo::object_groups< image2d<unsigned> >& groups, && (unsigned) abs(p1[0] - p2[0]) < 3 * max_height // Reduced gap && (unsigned) abs(p1[1] - p2[1]) < 20) // Vertical proximity { - draw::line(ima_groups, p1, p2, literal::red); - balance[i] += 1; - break; + unsigned k = 1; + short separators = 0; + + while (k <= hlines.nelements() && separators < 2) + { + const box2d& s = hlines(k).bbox(); + + if (s.pmin()[1] <= b1.pmin()[1] && s.pmax()[1] >= b1.pmax()[1] + && s.pmax()[0] < b1.pmin()[0] + && s.pmin()[0] > b2.pmax()[0]) + ++separators; + + ++k; + } + + if (separators < 2) + { + draw::line(ima_groups, p1, p2, literal::red); + balance[i] += 1; + break; + } } } } @@ -165,9 +205,11 @@ void draw_links_bt(const scribo::object_groups< image2d<unsigned> >& groups, } // Draw horizontal links from left to right (green) +template<typename L> void draw_links_lr(const scribo::object_groups< image2d<unsigned> >& groups, image2d<value::rgb8>& ima_groups, - std::vector<short>& balance) + std::vector<short>& balance, + const scribo::component_set<L>& vlines) { for (unsigned i = 1; i <= groups.nelements(); ++i) { @@ -186,9 +228,27 @@ void draw_links_lr(const scribo::object_groups< image2d<unsigned> >& groups, || (b1.pmin()[0] > b2.pmin()[0] && b2.pmax()[0] > b1.pmin()[0])) // Boxes are aligned && abs(p1[0] - p2[0]) < 10) // Reduced gap { - draw::line(ima_groups, p1, p2, literal::green); - balance[i] += 1; - break; + unsigned k = 1; + short separators = 0; + + while (k <= vlines.nelements() && separators < 2) + { + const box2d& s = vlines(k).bbox(); + + if (s.pmin()[0] <= b1.pmin()[0] && s.pmax()[0] >= b1.pmax()[0] + && s.pmin()[1] > b1.pmax()[1] + && s.pmax()[1] < b2.pmin()[1]) + ++separators; + + ++k; + } + + if (separators < 2) + { + draw::line(ima_groups, p1, p2, literal::green); + balance[i] += 1; + break; + } } } } @@ -196,9 +256,11 @@ void draw_links_lr(const scribo::object_groups< image2d<unsigned> >& groups, } // Draw horizontal links from right to left (green) +template<typename L> void draw_links_rl(const scribo::object_groups< image2d<unsigned> >& groups, image2d<value::rgb8>& ima_groups, - std::vector<short>& balance) + std::vector<short>& balance, + const scribo::component_set<L>& vlines) { for (unsigned i = groups.nelements(); i > 0; --i) { @@ -217,9 +279,27 @@ void draw_links_rl(const scribo::object_groups< image2d<unsigned> >& groups, || (b1.pmin()[0] > b2.pmin()[0] && b2.pmax()[0] > b1.pmin()[0])) // Boxes are aligned && abs(p1[0] - p2[0]) < 10) // Reduced gap { - draw::line(ima_groups, p1, p2, literal::green); - balance[i] += 1; - break; + unsigned k = 1; + short separators = 0; + + while (k <= vlines.nelements() && separators < 2) + { + const box2d& s = vlines(k).bbox(); + + if (s.pmin()[0] <= b1.pmin()[0] && s.pmax()[0] >= b1.pmax()[0] + && s.pmax()[1] < b1.pmin()[1] + && s.pmin()[1] > b2.pmax()[1]) + ++separators; + + ++k; + } + + if (separators < 2) + { + draw::line(ima_groups, p1, p2, literal::green); + balance[i] += 1; + break; + } } } } @@ -358,12 +438,13 @@ int main(int argc, char** argv) std::vector<short> balance(groups.nelements(), 0); // Draw and count links - draw_links_tb(groups, ima_groups, balance, average_width); - draw_links_bt(groups, ima_groups, balance, average_width); - draw_links_lr(groups, ima_groups, balance); - draw_links_rl(groups, ima_groups, balance); + draw_links_tb(groups, ima_groups, balance, average_width, hlines); + draw_links_bt(groups, ima_groups, balance, average_width, hlines); + draw_links_lr(groups, ima_groups, balance, vlines); + draw_links_rl(groups, ima_groups, balance, vlines); - // Draw weighted boxes (red < orange < cyan < green) (useless ?) + // Draw weighted boxes (red < orange < cyan < green) + // 1 link < 2 links < 3 links < 3+ links for (unsigned i = 0; i < balance.size(); ++i) { if (balance[i] == 1) -- 1.7.2.5

10 years, 11 months

olena: olena-2.0-563-g2661c4d [ICDAR_13] Check old version XML output with GUI Annotator

by Anthony Seure

--- scribo/sandbox/icdar_13_table/Makefile | 3 ++- .../competition-entry-region-model.xsd | 0 scribo/sandbox/icdar_13_table/src/main.cc | 9 ++++++--- 3 files changed, 8 insertions(+), 4 deletions(-) rename scribo/sandbox/icdar_13_table/{ => originals}/competition-entry-region-model.xsd (100%) diff --git a/scribo/sandbox/icdar_13_table/Makefile b/scribo/sandbox/icdar_13_table/Makefile index 7c11bf1..baa9ae1 100644 --- a/scribo/sandbox/icdar_13_table/Makefile +++ b/scribo/sandbox/icdar_13_table/Makefile @@ -1,6 +1,7 @@ CCACHE=ccache CC=g++ CFLAGS=-Wall -Werror -O3 -DHAVE_TESSERACT_3 -DNDEBUG -DMLN_WO_GLOBAL_VARS +CFLAGS_OLD=-Wall -Werror -O3 -DHAVE_TESSERACT_3 -DNDEBUG CLIBS=-I../../../milena/ -I../../ -I/usr/include/poppler CLEAN=*.o output/* log final.xml @@ -15,7 +16,7 @@ table: $(CCACHE) $(CC) $(CFLAGS) $(CLIBS) $(SRC) -ltesseract -lpoppler-cpp -o $(OUTPUT) old: - $(CCACHE) $(CC) $(CFLAGS) $(CLIBS) $(SRC_OLD) -ltesseract -lpoppler-cpp -o $(OUTPUT_OLD) + $(CCACHE) $(CC) $(CFLAGS_OLD) $(CLIBS) $(SRC_OLD) -ltesseract -lpoppler-cpp -o $(OUTPUT_OLD) clean: rm -rf $(CLEAN) diff --git a/scribo/sandbox/icdar_13_table/competition-entry-region-model.xsd b/scribo/sandbox/icdar_13_table/originals/competition-entry-region-model.xsd similarity index 100% rename from scribo/sandbox/icdar_13_table/competition-entry-region-model.xsd rename to scribo/sandbox/icdar_13_table/originals/competition-entry-region-model.xsd diff --git a/scribo/sandbox/icdar_13_table/src/main.cc b/scribo/sandbox/icdar_13_table/src/main.cc index be394ba..3aa40c2 100644 --- a/scribo/sandbox/icdar_13_table/src/main.cc +++ b/scribo/sandbox/icdar_13_table/src/main.cc @@ -26,9 +26,12 @@ using namespace mln; -void start_xml(std::ofstream& xml, const char* name, const char* pdf) +void start_xml(std::ofstream& xml, const char* pdf) { - xml.open(name); + //std::ostringstream name; + //name << pdf << "-reg-result.xml"; + //xml.open((name.str()).c_str()); + xml.open("us-005-reg-result.xml"); xml << "<?xml version\"1.0\" encoding=\"UTF-8\"?>" << std::endl << "<document filename='" << pdf << "'>" << std::endl; } @@ -212,7 +215,7 @@ int main(int argc, char** argv) // Loading and binarization std::ofstream xml; - start_xml(xml, "final.xml", argv[1]); + start_xml(xml, argv[1]); //io::ppm::load(original, argv[1]); util::array< image2d<value::rgb8> > pdf; -- 1.7.2.5

10 years, 11 months

olena: olena-2.0-562-g8241f48 [ICDAR_13] Update XML output + Add XML reference

by Anthony Seure

--- scribo/sandbox/icdar_13_table/Makefile | 2 +- .../competition-entry-region-model.xsd | 45 +++++++++++++++++++ scribo/sandbox/icdar_13_table/src/new.cc | 5 +-- scribo/sandbox/icdar_13_table/src/xml.cc | 47 +++++++++++++------ scribo/sandbox/icdar_13_table/src/xml.hh | 9 +++- 5 files changed, 86 insertions(+), 22 deletions(-) create mode 100644 scribo/sandbox/icdar_13_table/competition-entry-region-model.xsd diff --git a/scribo/sandbox/icdar_13_table/Makefile b/scribo/sandbox/icdar_13_table/Makefile index 8d0cd1a..7c11bf1 100644 --- a/scribo/sandbox/icdar_13_table/Makefile +++ b/scribo/sandbox/icdar_13_table/Makefile @@ -1,6 +1,6 @@ CCACHE=ccache CC=g++ -CFLAGS=-Wall -Werror -O3 -DHAVE_TESSERACT_3 -DNDEBUG +CFLAGS=-Wall -Werror -O3 -DHAVE_TESSERACT_3 -DNDEBUG -DMLN_WO_GLOBAL_VARS CLIBS=-I../../../milena/ -I../../ -I/usr/include/poppler CLEAN=*.o output/* log final.xml diff --git a/scribo/sandbox/icdar_13_table/competition-entry-region-model.xsd b/scribo/sandbox/icdar_13_table/competition-entry-region-model.xsd new file mode 100644 index 0000000..2e0d0d4 --- /dev/null +++ b/scribo/sandbox/icdar_13_table/competition-entry-region-model.xsd @@ -0,0 +1,45 @@ +<?xml version="1.0" encoding="utf-8"?> +<xsd:schema attributeFormDefault="unqualified" elementFormDefault="qualified" version="1.0" + xmlns:xsd="http://www.w3.org/2001/XMLSchema" > + <xsd:element name="document"> + <xsd:complexType> + <xsd:sequence> + <xsd:element minOccurs="0" maxOccurs="unbounded" name="table">  + <xsd:complexType> + <xsd:sequence> + <xsd:element name="region" maxOccurs="unbounded" nillable="false">  + <xsd:complexType> + <xsd:sequence> + <xsd:element name="instruction" minOccurs="0" maxOccurs="unbounded">  + <xsd:complexType> + <xsd:attribute name="instr-id" type="xsd:integer" use="required"/> + <xsd:attribute name="subinstr-id" type="xsd:integer"/> +  + </xsd:complexType> + </xsd:element> + <xsd:element name="bounding-box">  + <xsd:complexType> + <xsd:attribute name="x1" type="xsd:integer" use="required"/> + <xsd:attribute name="y1" type="xsd:integer" use="required"/> + <xsd:attribute name="x2" type="xsd:integer" use="required"/> + <xsd:attribute name="y2" type="xsd:integer" use="required"/> + </xsd:complexType> + </xsd:element> + </xsd:sequence> + <xsd:attribute name="id" type="xsd:nonNegativeInteger" use="required"/> + <xsd:attribute name="page" type="xsd:positiveInteger" use="required"/> + </xsd:complexType> + </xsd:element> + </xsd:sequence> + <xsd:attribute name="id" type="xsd:nonNegativeInteger" use="required"/> + </xsd:complexType> + </xsd:element> + </xsd:sequence> + <xsd:attribute name="filename" type="xsd:string" use="required"/> + </xsd:complexType> + </xsd:element> +</xsd:schema> diff --git a/scribo/sandbox/icdar_13_table/src/new.cc b/scribo/sandbox/icdar_13_table/src/new.cc index 95bf575..9f05030 100644 --- a/scribo/sandbox/icdar_13_table/src/new.cc +++ b/scribo/sandbox/icdar_13_table/src/new.cc @@ -1,3 +1,4 @@ +#undef MLN_WO_GLOBAL_VARS #include "xml.hh" // INCLUDES OLENA @@ -231,7 +232,6 @@ int main(int argc, char** argv) typedef value::label_16 V; typedef image2d<V> L; - //std::ofstream xml; std::ostringstream path; image2d<value::rgb8> original, ima_links, ima_groups, ima_valid; image2d<value::int_u8> filtered; @@ -241,7 +241,6 @@ int main(int argc, char** argv) unsigned dpi = 72; // Loading and binarization - //start_xml(xml, "final.xml", argv[1]); XML* xml = new XML("final.xml", argv[1]); util::array< image2d<value::rgb8> > pdf; @@ -395,8 +394,6 @@ int main(int argc, char** argv) write_image(ima_valid, "valid", page, number, path); } - - //end_xml(xml); delete xml; return 0; diff --git a/scribo/sandbox/icdar_13_table/src/xml.cc b/scribo/sandbox/icdar_13_table/src/xml.cc index 76fed84..86e66e2 100644 --- a/scribo/sandbox/icdar_13_table/src/xml.cc +++ b/scribo/sandbox/icdar_13_table/src/xml.cc @@ -1,31 +1,48 @@ #include "xml.hh" -XML::XML(const char* name, const char* pdf) - : _name(name), _pdf(pdf) +XML::XML(const char* name, + const char* pdf) + : _name(name), _pdf(pdf), _table(0), _region(0), _first_time(true) { _xml.open(_name); _xml << "<?xml version\"1.0\" encoding=\"UTF-8\"?>" << std::endl - << "<document filename='" << _pdf << "'>" << std::endl; + << "<document filename='" << _pdf << "'>" << std::endl; } -XML::~XML() +XML::~XML(void) { _xml << "</document>" << std::endl; _xml.close(); } -void XML::write_table(const point2d& start, const point2d& end) +void XML::table(const point2d& start, + const point2d& end, + const unsigned page, + const bool connect) { - static unsigned table = 0; - static unsigned region = 0; - static unsigned page = 1; + if (_first_time) + { + _xml << "\t<table id='" << _table << "'>" << std::endl; + ++_table; + _first_time = false; + } + else + { + if (!connect) + { + _xml << "\t</table>" << std::endl; + _xml << "\t<table id='" << _table << "'>" << std::endl; + _region = 0; + ++_table; + } + } - _xml << "\t<table id='" << table << "'>" << std::endl - << "\t\t<region id='" << region << "' page='" << page << "'>" << std::endl - << "\t\t<bounding-box x1='" << start[1] << "' y1='" << start[0] << "' " - << "x2='" << end[1] << "' y2='" << end[0] << "'/>" << std::endl - << "\t\t</region>" << std::endl - << "\t</table>" << std::endl; + _xml << "\t\t<region id='" << _region << "' page='" << page + 1 << "'>" << std::endl + << "\t\t\t<bounding-box x1='" << start[1] + << "' y1='" << start[0] + << "' x2='" << end[1] + << "' y2='" << end[0] << "'/>" << std::endl + << "\t\t</region>" << std::endl; - ++table; + ++_region; } diff --git a/scribo/sandbox/icdar_13_table/src/xml.hh b/scribo/sandbox/icdar_13_table/src/xml.hh index 397d585..b4cef29 100644 --- a/scribo/sandbox/icdar_13_table/src/xml.hh +++ b/scribo/sandbox/icdar_13_table/src/xml.hh @@ -1,6 +1,5 @@ #ifndef XML_HH # define XML_HH -# define MLN_WO_GLOBAL_VARS # include <iostream> # include <fstream> @@ -13,12 +12,18 @@ class XML public: XML(const char* name, const char* pdf); ~XML(); - void write_table(const point2d& start, const point2d& end); + void table(const point2d& start, + const point2d& end, + const unsigned page, + const bool connect); private: std::ofstream _xml; const char* _name; const char* _pdf; + unsigned _table; + unsigned _region; + bool _first_time; }; #endif /* !XML_HH */ -- 1.7.2.5

10 years, 11 months

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

Olena-patches June 2013