https://svn.lrde.epita.fr/svn/oln/trunk/milena/sandbox Index: ChangeLog from Thierry Geraud <thierry.geraud@lrde.epita.fr> Add ICDAR-related code. * theo/icdar/dibco/wst-based.sh: New. * theo/icdar/hsc/boxes.cc: New. * theo/icdar/hsc/main.cc: New. dibco/wst-based.sh | 8 + hsc/boxes.cc | 102 ++++++++++++++ hsc/main.cc | 385 +++++++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 495 insertions(+) Index: theo/icdar/dibco/wst-based.sh --- theo/icdar/dibco/wst-based.sh (revision 0) +++ theo/icdar/dibco/wst-based.sh (revision 0) @@ -0,0 +1,8 @@ +#! /bin/zsh + +./+bin/elementary_gradient $1 1 tmp_grad.pgm +./+bin/closing_volume tmp_grad.pgm 500 tmp_clo.pgm +./+bin/watershed_flooding tmp_clo.pgm tmp_wst.pgm +./+bin/watershed_superpose $1 tmp_wst.pgm ${1%pgm}ppm + +rm tmp_grad.pgm tmp_clo.pgm tmp_wst.pgm Property changes on: theo/icdar/dibco/wst-based.sh ___________________________________________________________________ Added: svn:executable + * Index: theo/icdar/hsc/boxes.cc --- theo/icdar/hsc/boxes.cc (revision 0) +++ theo/icdar/hsc/boxes.cc (revision 0) @@ -0,0 +1,102 @@ +// Copyright (C) 2009 EPITA Research and Development Laboratory +// +// This file is part of the Olena Library. This library is free +// software; you can redistribute it and/or modify it under the terms +// of the GNU General Public License version 2 as published by the +// Free Software Foundation. +// +// This library is distributed in the hope that it will be useful, +// but WITHOUT ANY WARRANTY; without even the implied warranty of +// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +// General Public License for more details. +// +// You should have received a copy of the GNU General Public License +// along with this library; see the file COPYING. If not, write to +// the Free Software Foundation, 51 Franklin Street, Fifth Floor, +// Boston, MA 02111-1307, USA. +// +// As a special exception, you may use this file as part of a free +// software library without restriction. Specifically, if other files +// instantiate templates or use macros or inline functions from this +// file, or you compile this file and link it with other files to +// produce an executable, this file does not by itself cause the +// resulting executable to be covered by the GNU General Public +// License. This exception does not however invalidate any other +// reasons why the executable file might be covered by the GNU General +// Public License. + +#include <iostream> + +#include <mln/essential/2d.hh> +#include <mln/debug/colorize.hh> +#include <mln/debug/println.hh> + +#include <scribo/text/extract_bboxes.hh> +#include <scribo/text/grouping/group_with_single_left_link.hh> +#include <scribo/text/grouping/group_with_single_right_link.hh> +#include <scribo/debug/save_linked_textbboxes_image.hh> +#include <scribo/text/grouping/group_from_double_link.hh> +#include <scribo/filter/small_components.hh> + +#include <scribo/debug/save_textbboxes_image.hh> +#include <scribo/make/debug_filename.hh> + +#include <mln/logical/not.hh> +#include <mln/io/dump/save.hh> + + +int usage(const char *name) +{ + std::cout << "Usage: " << name << " input.pbm output.dump" << std::endl; + return 1; +} + + +int main(int argc, char* argv[]) +{ + using namespace scribo; + using namespace mln; + + if (argc != 3) + return usage(argv[0]); + + scribo::make::internal::debug_filename_prefix = "extract_text_double_link"; + + image2d<bool> input; + io::pbm::load(input, argv[1]); + + logical::not_inplace(input); + + value::label_16 nbboxes; + scribo::util::text<image2d<value::label_16> > + text = text::extract_bboxes(input, c8(), nbboxes); + + text = filter::small_components(text,4); + + mln::util::array<unsigned> + left_link = text::grouping::group_with_single_left_link(text, 100), + right_link = text::grouping::group_with_single_right_link(text, 100); + + std::cout << "BEFORE - nbboxes = " << nbboxes << std::endl; + + + scribo::debug::save_linked_textbboxes_image(input, + text, left_link, right_link, + literal::red, literal::cyan, literal::yellow, + literal::green, + scribo::make::debug_filename("links.ppm")); + + // Validation. + scribo::util::text<image2d<value::label_16> > grouped_text + = text::grouping::group_from_double_link(text, left_link, right_link); + + std::cout << "AFTER double grouping - nbboxes = " << grouped_text.bboxes().nelements() << std::endl; + + io::dump::save(grouped_text.label_image(), argv[2]); + + scribo::debug::save_textbboxes_image(input, grouped_text.bboxes(), + literal::red, + scribo::make::debug_filename("boxes.ppm")); + +} + Index: theo/icdar/hsc/main.cc --- theo/icdar/hsc/main.cc (revision 0) +++ theo/icdar/hsc/main.cc (revision 0) @@ -0,0 +1,385 @@ +#include <set> +#include <vector> + +#include <mln/core/image/image2d.hh> +#include <mln/core/image/image_if.hh> +#include <mln/core/alias/neighb2d.hh> + +#include <mln/pw/all.hh> +#include <mln/data/fill.hh> +#include <mln/level/saturate.hh> +#include <mln/level/convert.hh> +#include <mln/arith/revert.hh> + +#include <mln/value/int_u8.hh> +#include <mln/value/rgb8.hh> +#include <mln/literal/colors.hh> +#include <mln/debug/colorize.hh> + +#include <mln/io/pbm/load.hh> +#include <mln/io/pbm/save.hh> +#include <mln/io/pgm/save.hh> +#include <mln/io/ppm/save.hh> + +#include <mln/world/binary_2d/subsample.hh> + +#define MLN_FLOAT double +#include <sandbox/theo/exec/gaussian_directional_2d.hh> + +#include <mln/morpho/closing/structural.hh> +#include <mln/morpho/watershed/flooding.hh> +#include <mln/win/rectangle2d.hh> + +#include <mln/accu/stat/variance.hh> + + + +namespace mln +{ + + + template <typename L> + inline + L find_root_(std::vector<L>& parent, L l) + { + if (parent[l] == l) + return l; + return parent[l] = find_root_(parent, parent[l]); + } + + + + template <typename L> + image2d<L> + merge(const image2d<value::int_u8>& ima, + const image2d<L>& ws, + L n, + unsigned n_min_stats) + { + typedef accu::stat::variance<float> A; + std::vector<A> + vvar(n + 1), + hvar(n + 1); + + mln_piter(box2d) p(ima.domain()); + for_all(p) + if (ima(p) != 255 && ws(p) != 0) + vvar[ws(p)].take(255 - ima(p), p.row()); + + + image2d<bool> + radj(n + 1, n + 1), + ladj(n + 1, n + 1); + data::fill(radj, false); + data::fill(ladj, false); + + for_all(p) if (ws(p) == 0) + { + if (ws(p + left) == 0 || ws(p + right) == 0) + // ws is horizontal => forget it + continue; + L l1 = ws(p + left), + l2 = ws(p + right); + if (l1 == 0 || l2 == 0 || l2 == l1) + continue; + radj.at_(l1, l2) = true; + ladj.at_(l2, l1) = true; + } + + std::vector<L> + best_right(n + 1, 0), + best_left(n + 1, 0); + L best_l2; + float best_score; + + for (L l1 = 1; l1 <= n; ++l1) + { + if (! vvar[l1].is_valid() || vvar[l1].n_items() < n_min_stats) + // Non significative stats. + continue; + + + // Look right. + +# ifdef LOG + std::cout << l1 << ": "; +# endif + + best_l2 = 0; + best_score; + + for (L l2 = 1; l2 <= n; ++l2) + { + if (! radj.at_(l1, l2)) // l1 -> l2 + continue; + + if (! vvar[l2].is_valid() || vvar[l2].n_items() < n_min_stats) + // Non significative stats. + continue; + +# ifdef LOG + std::cout << l2 << ' '; +# endif + + float vm, vM, sm, sM; + if (vvar[l1].mean() < vvar[l2].mean()) + { + vm = vvar[l1].mean(); + sm = vvar[l1].standard_deviation(); + vM = vvar[l2].mean(); + sM = vvar[l2].standard_deviation(); + } + else + { + vm = vvar[l2].mean(); + sm = vvar[l2].standard_deviation(); + vM = vvar[l1].mean(); + sM = vvar[l1].standard_deviation(); + } + + float score = (vm - sm) - (vM - sM); +# ifdef LOG + std::cout << '(' << score << "), "; +# endif + if (best_l2 == 0 || score > best_score) + { + best_score = score; + best_l2 = l2; + } + } + + if (best_l2 != 0) + { + best_right[l1] = best_l2; +# ifdef LOG + std::cout << " ... " << l1 << " -> " << best_l2 << std::endl; +# endif + } + + // Look left. + +# ifdef LOG + std::cout << l1 << ": "; +# endif + + best_l2 = 0; + best_score; + + for (L l2 = 1; l2 <= n; ++l2) + { + if (! ladj.at_(l1, l2)) // l2 <- l1 + continue; + + if (! vvar[l2].is_valid() || vvar[l2].n_items() < n_min_stats) + // Non significative stats. + continue; + +# ifdef LOG + std::cout << l2 << ' '; +# endif + + float vm, vM, sm, sM; + if (vvar[l1].mean() < vvar[l2].mean()) + { + vm = vvar[l1].mean(); + sm = vvar[l1].standard_deviation(); + vM = vvar[l2].mean(); + sM = vvar[l2].standard_deviation(); + } + else + { + vm = vvar[l2].mean(); + sm = vvar[l2].standard_deviation(); + vM = vvar[l1].mean(); + sM = vvar[l1].standard_deviation(); + } + + float score = (vm - sm) - (vM - sM); +# ifdef LOG + std::cout << '(' << score << "), "; +# endif + + if (best_l2 == 0 || score > best_score) + { + best_score = score; + best_l2 = l2; + } + } + + if (best_l2 != 0) + { + best_left[l1] = best_l2; +# ifdef LOG + std::cout << " ... " << best_l2 << " <- " << l1 << std::endl; +# endif + } + + } + + + std::vector<L> parent(n + 1); + for (L l = 0; l <= n; ++l) + parent[l] = l; + + for (L l1 = 1; l1 <= n; ++l1) + { + L l2 = best_right[l1]; + if (l2 == 0) + continue; + if (best_left[l2] == l1) + { + // cross-validation + L l1r = find_root_(parent, l1), + l2r = find_root_(parent, l2); + if (l2r == l1r) + continue; // already merged + if (l1r < l2r) + parent[l1r] = l2r; + else + parent[l2r] = l1r; + } + } + + + for (L l = 1; l <= n; ++l) + parent[l] = find_root_(parent, l); + + + image2d<L> out(ima.domain()); + for_all(p) + out(p) = parent[ws(p)]; + + // io::pgm::save(out, "tmp_out.pgm"); + + return out; + + } // end of 'merge' + + + +} // ! mln + + + + +void usage(char* argv[]) +{ + std::cerr << "usage: " << argv[0] << " input.pbm output.pgm [output.ppm]" << std::endl + << " ICDAR'2009: HSC." << std::endl; + std::abort(); +} + + + +int main(int argc, char* argv[]) +{ + using namespace mln; + using value::int_u8; + using value::rgb8; + + if (argc != 3 && argc != 4) + usage(argv); + + + // Parameters. + + const unsigned + subsampling_factor = 4, + height = 5, + width = 25, + n_min_stats = 1000; + const float + h_sigma = 31, + v_sigma = 1.3; + + // end of Parameters. + + + trace::entering("main"); + + image2d<bool> input; + io::pbm::load(input, argv[1]); + + // Sub-sampling. + image2d<int_u8> + small = world::binary_2d::subsample(input, subsampling_factor), + fuzzy, clo, + ws, + spc; + + +# ifdef LOG + io::pgm::save(small, "tmp_small.pgm"); +# endif + + + // Fuzzifying. + { + image2d<MLN_FLOAT> temp(small.domain()), out; + data::fill(temp, small); + + out = linear::gaussian_directional_2d(temp, 1, h_sigma, 0); + out = linear::gaussian_directional_2d(out, 0, v_sigma, 0); + + fuzzy = level::saturate(int_u8(), out); + +# ifdef LOG + io::pgm::save(fuzzy, "tmp_fuzzy.pgm"); +# endif + } + + + clo = morpho::closing::structural(fuzzy, win::rectangle2d(height, width)); + + int_u8 n_basins; + ws = morpho::watershed::flooding(clo, c4(), n_basins); + + + spc = merge(small, ws, n_basins, n_min_stats); // Merging! + + +# ifdef LOG + + { + io::ppm::save(debug::colorize(rgb8(), spc, n_basins), + "tmp_spc.ppm"); + + io::pgm::save(ws, "tmp_ws.pgm"); + + image2d<rgb8> cool = level::convert(rgb8(), small); + data::fill((cool | (pw::value(ws) == pw::cst(0))).rw(), + literal::red); + io::ppm::save(cool, "tmp_ws.ppm"); + } +# endif + + + // Outputing. + { + image2d<int_u8> output(input.domain()); + + image2d<rgb8> + cool = debug::colorize(rgb8(), spc, n_basins), + lab(input.domain()); + + mln_piter_(box2d) p(input.domain()); + for_all(p) + if (input(p)) + { + output(p) = 0; + lab(p) = literal::black; + } + else + { + output(p) = spc.at_(p.row() / subsampling_factor, p.col() / subsampling_factor); + lab(p) = cool.at_(p.row() / subsampling_factor, p.col() / subsampling_factor); + } + + io::pgm::save(output, argv[2]); + if (argc == 4) + io::ppm::save(lab, argv[3]); + } + + + trace::exiting("main"); +}