Olena-patches

olena-patches@lrde.epita.fr

9625 discussions

branch origin/julien_temp_paragraphs deleted: last-svn-commit-842-g4ef87a3

by git＠lrde.epita.fr

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "Olena, a generic and efficient image processing platform". The branch origin/julien_temp_paragraphs has been deleted was 4ef87a32cc19fbb7e5bd622bd9bd79cdd58c3cb9 ----------------------------------------------------------------------- 4ef87a32cc19fbb7e5bd622bd9bd79cdd58c3cb9 icdar : paragraphs extraction ----------------------------------------------------------------------- hooks/post-receive -- Olena, a generic and efficient image processing platform

13 years, 8 months

branch julien_temp_paragraphs updated: last-svn-commit-842-g4ef87a3

by git＠lrde.epita.fr

13 years, 8 months

branch origin/julien_temp_paragraphs created: last-svn-commit-842-g4ef87a3

by git＠lrde.epita.fr

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "Olena, a generic and efficient image processing platform". The branch origin/julien_temp_paragraphs has been created at 4ef87a32cc19fbb7e5bd622bd9bd79cdd58c3cb9 (commit) - Log ----------------------------------------------------------------- 4ef87a3 icdar : paragraphs extraction ----------------------------------------------------------------------- hooks/post-receive -- Olena, a generic and efficient image processing platform

13 years, 8 months

last-svn-commit-842-g4ef87a3 icdar : paragraphs extraction

by Julien Marquegnies

--- scribo/scribo/core/line_info.hh | 18 +- scribo/scribo/core/stats.hh | 16 +- scribo/scribo/text/merging.hh | 16 +- scribo/scribo/text/paragraphs.hh | 462 ++++++++++++++++++++++++++++++------- 4 files changed, 402 insertions(+), 110 deletions(-) diff --git a/scribo/scribo/core/line_info.hh b/scribo/scribo/core/line_info.hh index a798e9b..fa4051d 100644 --- a/scribo/scribo/core/line_info.hh +++ b/scribo/scribo/core/line_info.hh @@ -1006,21 +1006,22 @@ namespace scribo const int dy = bb1.pmax().row() - bb2.pmax().row(); // The two characters must be distinct - if (space < 0) - return false; + // if (space < 0) + // return false; if (// Approximately the same width - ((std::max(w1, w2) / std::min(w1, w2)) > 1.1f || + ((std::max(w1, w2) / std::min(w1, w2)) > 1.3 || // One character must not be smaller than the space between // the two characters (w1 < space || w2 < space)) // If the two characters have a different width they must also // have a different height - && not (std::max(h1, h2) / std::min(h1, h2) <= 1.5f)) + && not (std::max(h1, h2) / std::min(h1, h2) <= 1.7f)) return false; // Approximately aligned on baseline - if (std::abs(dy) > 10) + if (std::abs(dy) > 10 && + not (std::max(h1, h2) / std::min(h1, h2) <= 1.7f)) return false; return true; @@ -1055,6 +1056,10 @@ namespace scribo float min_base = 0.0f; const unsigned clusters_b_nelements = clusters_b.nelements(); + if (clusters_b_nelements >= 3) + return data_->baseline_clusters_.mean(); + + for (unsigned i = 0; i < clusters_b_nelements; ++i) { const unsigned clusters_b_i_nelements = clusters_b[i].nelements(); @@ -1092,6 +1097,9 @@ namespace scribo float max_mean = 0.0f; const unsigned clusters_m_nelements = clusters_m.nelements(); + if (clusters_m_nelements >= 3) + return data_->meanline_clusters_.mean(); + for (unsigned i = 0; i < clusters_m_nelements; ++i) { const unsigned clusters_m_i_nelements = clusters_m[i].nelements(); diff --git a/scribo/scribo/core/stats.hh b/scribo/scribo/core/stats.hh index 43cc3d6..3801545 100644 --- a/scribo/scribo/core/stats.hh +++ b/scribo/scribo/core/stats.hh @@ -250,7 +250,7 @@ private: std::vector< unsigned > clusters; unsigned cluster_index = 1; - clusters.reserve(data_.nelements()); + clusters.resize(data_.nelements()); if (not data_sorted_) { @@ -261,7 +261,7 @@ private: unsigned i = 0; const unsigned nelements = data_.nelements(); - clusters.push_back(cluster_index); + clusters[0] = cluster_index; const T std = data_.standard_deviation(); for (i = 1; i < nelements - 1; ++i) @@ -273,16 +273,14 @@ private: && left_distance <= std)) ++cluster_index; - clusters.push_back(cluster_index); + clusters[i] = cluster_index; } - if (nelements > 1) - { - if (data_[i] - data_[i - 1] > std) - ++cluster_index; + if (nelements > 1 + && data_[i] - data_[i - 1] > std) + ++cluster_index; - clusters.push_back(cluster_index); - } + clusters[i] = cluster_index; clusters_.clear(); clusters_.reserve(cluster_index); diff --git a/scribo/scribo/text/merging.hh b/scribo/scribo/text/merging.hh index c94f9f5..9b4db56 100644 --- a/scribo/scribo/text/merging.hh +++ b/scribo/scribo/text/merging.hh @@ -460,8 +460,8 @@ namespace scribo if (l_cur_height < l_ted_x_height && l_cur_height > 0.05f * l_ted_x_height && float(l_cur_width) / float(l_cur.card()) < l_ted.char_width() - && dx < l_ted_cw - && l_cur_pmin.row() < l_ted_pmax.row()) + && dx < 2 * l_ted_cw + && l_cur_pmin.row() < l_ted.baseline()) { l_cur.update_type(line::Punctuation); return true; @@ -740,11 +740,12 @@ namespace scribo // vertically aligned // Obviously no separators between the two lines if ((l_info.card() <= 5 || - (std::abs(l_info.baseline() - mc_info.baseline()) < 5 - && std::abs(l_info.meanline() - mc_info.meanline()) < 5)) - && dx < l_ted_cw && dy < 0 - && not (l_info.holder().components().has_separators() - && between_separators(l_info, mc_info))) + (std::abs(l_info.baseline() - mc_info.baseline()) + < 5 && std::abs(l_info.meanline() - + mc_info.meanline()) < 5)) + && dx < l_ted_cw && dy < 0 + && not (l_info.holder().components().has_separators() + && between_separators(l_info, mc_info))) l = do_union(lines, l, mc, parent); // } @@ -1046,7 +1047,6 @@ namespace scribo // ts = t.stop(); // std::cout << "time " << ts << std::endl; - lines.force_stats_update(); return lines; diff --git a/scribo/scribo/text/paragraphs.hh b/scribo/scribo/text/paragraphs.hh index 5cb253e..beaea0c 100644 --- a/scribo/scribo/text/paragraphs.hh +++ b/scribo/scribo/text/paragraphs.hh @@ -47,11 +47,123 @@ namespace scribo return x; } + + template <typename T> + inline + void + set_root(util::array<T>& parent, unsigned x, const unsigned root) + { + while (parent(x) != x && parent(x) != root) + { + const unsigned tmp = parent(x); + x = parent(x); + parent(tmp) = root; + } + + parent(x) = root; + } } namespace filter { + template <typename L> + inline + bool + between_horizontal_separator(const scribo::line_info<L>& l1, + const scribo::line_info<L>& l2) + { + // No separators found in image. + mln_precondition(l1.holder().components().has_separators()); + + const box2d& l1_bbox = l1.bbox(); + const box2d& l2_bbox = l2.bbox(); + + unsigned + row1 = l1_bbox.pcenter().row(), + row2 = l2_bbox.pcenter().row(); + const mln_ch_value(L, bool)& + separators = l1.holder().components().separators(); + + unsigned row; + unsigned col_ptr; + unsigned left_col_ptr; + unsigned right_col_ptr; + unsigned end; + + if (row1 < row2) + { + row1 = l1_bbox.pmax().row(); + row2 = l2_bbox.pmin().row(); + + const unsigned quarter = + ((l1_bbox.pcenter().col() - l1_bbox.pmin().col()) >> 2); + + row = l1_bbox.pcenter().row(); + col_ptr = l1_bbox.pcenter().col(); + left_col_ptr = l1_bbox.pmin().col() + quarter; + right_col_ptr = l1_bbox.pmax().col() - quarter; + end = row2; + } + else + { + row2 = l2_bbox.pmax().row(); + row1 = l1_bbox.pmin().row(); + + const unsigned quarter = + ((l2_bbox.pcenter().col() - l2_bbox.pmin().col()) >> 2); + + row = l2_bbox.pcenter().row(); + col_ptr = l2_bbox.pcenter().col(); + left_col_ptr = l2_bbox.pmin().col() + quarter; + right_col_ptr = l2_bbox.pmax().col() - quarter; + end = row1; + } + + // If sep_ptr is true, then a separator is reached. + while (row < end) + { + ++row; + if (separators.at_(row, col_ptr) + || separators.at_(row, left_col_ptr) + || separators.at_(row, right_col_ptr)) + return true; + } + + return false; + } + + + template <typename L> + inline + bool may_have_another_left_link(const util::array<value::int_u16>& right, + const value::int_u16 index, + const value::int_u16 current_line, + const util::array< line_info<L> >& lines) + { + const line_info<L>& l = lines(current_line); + const point2d& pmin = l.bbox().pmin(); + const unsigned x1 = l.x_height(); + + for (unsigned i = 0; i < right.nelements(); ++i) + { + if (i != index && right(i) == index) + { + const line_info<L>& l_info = lines(i); + const unsigned x2 = l_info.x_height(); + + const float delta_max = 0.5f * std::min(x1, x2); + + if (l_info.bbox().pmin().col() < pmin.col() + && std::abs(l.baseline() - l_info.baseline()) < delta_max + ) + return true; + } + } + + return false; + } + //--------------------------------------------------------------------- // This method aims to cut the links between lines that do not fit the // different criteria @@ -60,7 +172,7 @@ namespace scribo template <typename L> inline void paragraph_links(const util::array<value::int_u16>& left, - const util::array<value::int_u16>& right, + util::array<value::int_u16>& right, util::array<value::int_u16>& output, const util::array< line_info<L> >& lines, const image2d<bool>& input) @@ -81,9 +193,28 @@ namespace scribo { // Neighbors - const value::int_u16 left_nbh = output(l); - const value::int_u16 right_nbh = right(l); - const value::int_u16 lol_nbh = output(left_nbh); + value::int_u16 left_nbh = output(l); + value::int_u16 right_nbh = right(l); + value::int_u16 lol_nbh = output(left_nbh); + + const line_info<L>& left_line = lines(left_nbh); + const line_info<L>& current_line = lines(l); + const line_info<L>& right_line = lines(right_nbh); + + if (right_line.holder().components().has_separators() && + between_horizontal_separator(right_line, current_line)) + { + output(right_nbh) = right_nbh; + right_nbh = l; + } + if (current_line.holder().components().has_separators() && + between_horizontal_separator(current_line, left_line)) + { + output(l) = l; + left_nbh = l; + lol_nbh = l; + } + // Line features const float x_height = lines(l).x_height(); @@ -110,6 +241,7 @@ namespace scribo // Maximal x variation to consider two lines vertically aligned const int delta_alignment = cline_cw; + // Checks the baseline distances of the two neighbors { // Current line baseline @@ -149,7 +281,7 @@ namespace scribo // and its right neighbor if (right_distance > 1.4f * ror_distance && std::max(ror_x_height, right_x_height) < - 1.2f * std::min(ror_x_height, right_x_height) + 1.4f * std::min(ror_x_height, right_x_height) && output(right_nbh) == l) { output(right_nbh) = right_nbh; @@ -182,7 +314,7 @@ namespace scribo // Condition to cut the link between the current line and // its right neighbor - if ((max_x_height > min_x_height * 1.2f) && + if ((max_x_height > min_x_height * 1.4f) && !(max_char_width <= 1.2f * min_char_width)) { if (output(right_nbh) == l) @@ -218,7 +350,7 @@ namespace scribo // and its left neighbor if (left_distance > 1.4f * lol_distance && std::max(lol_x_height, left_x_height) < - 1.2f * std::min(lol_x_height, left_x_height)) + 1.4f * std::min(lol_x_height, left_x_height)) { output(l) = l; continue; @@ -250,7 +382,7 @@ namespace scribo // Condition to cut the link between the current line and // its left neighbor - if ((max_x_height > min_x_height * 1.2f) && + if ((max_x_height > min_x_height * 1.4f) && !(max_char_width <= 1.2f * min_char_width)) { output(l) = l; @@ -262,27 +394,28 @@ namespace scribo continue; } // The current line has at least one left and one right neighbor - else // if (delta_baseline_max >= delta_baseline_min) + else // if (delta_baseline_max >= 1.1f * delta_baseline_min) { // Distance between the left and the current line const float left_distance = - lines(left_nbh).meanline() - lines(l).baseline(); + left_line_bbox.pcenter().row() - current_line_bbox.pcenter().row(); // Distance between the right and the current line const float right_distance = - lines(l).meanline() - lines(right_nbh).baseline(); + current_line_bbox.pcenter().row() - right_line_bbox.pcenter().row(); // If the left line is too far compared to the right one // we cut the link with it - if (left_distance > 1.2f * right_distance - && std::max(x_height, left_x_height) > 1.2f * std::min(x_height, left_x_height)) + if (left_distance > 1.5f * right_distance + && std::max(x_height, left_x_height) > 1.2f * std::min(x_height, left_x_height) + ) { output(l) = l; continue; } // If the right line is too far compared to the left one // we cut the link with it - else if (right_distance > 1.2f * left_distance - && std::max(x_height, right_x_height) > 1.2f * std::min(x_height, right_x_height) + else if (right_distance > 1.5f * left_distance + && std::max(x_height, right_x_height) >= 1.2f * std::min(x_height, right_x_height) && output(right_nbh) == l) { output(right_nbh) = right_nbh; @@ -301,7 +434,7 @@ namespace scribo const float min_x_height = std::min(x_height, left_x_height); const float max_x_height = std::max(x_height, left_x_height); - if ((max_x_height > min_x_height * 1.2f) && + if ((max_x_height > min_x_height * 1.4f) && !(cw_max <= 1.2f * cw_min)) { output(l) = l; @@ -314,7 +447,7 @@ namespace scribo const float cw_max = std::max(rline_cw, cline_cw); const float cw_min = std::min(rline_cw, cline_cw); - if ((max_x_height > min_x_height * 1.2f) + if ((max_x_height > min_x_height * 1.4f) && !(cw_max <= 1.2f * cw_min) && output(right_nbh) == l) { @@ -330,7 +463,7 @@ namespace scribo const float min_x_height = std::min(x_height, right_x_height); const float max_x_height = std::max(x_height, right_x_height); - if ((max_x_height > min_x_height * 1.2f) + if ((max_x_height > min_x_height * 1.4f) && !(cw_max <= 1.2f * cw_min) && output(right_nbh) == l) { @@ -344,7 +477,7 @@ namespace scribo const float cw_max = std::max(lline_cw, cline_cw); const float cw_min = std::min(lline_cw, cline_cw); - if ((max_x_height > min_x_height * 1.2f) + if ((max_x_height > min_x_height * 1.4f) && !(cw_max <= 1.2f * cw_min)) { output(l) = l; @@ -363,51 +496,6 @@ namespace scribo //----------------------------------------------------------------------------- // ___________________________ // |___________________________| -// ________________________ -// |________________________| -// ___________________________ -// |___________________________| -// ___________________________ -// |___________________________| -// -// Simple case : paragraphs are justified on the left. We try to find any -// indentation like above. -// -//----------------------------------------------------------------------------- - - { - // Check if the current line neighbors are aligned - bool left_right_aligned = false; - bool left_lol_aligned = false; - const int dx_lr = std::abs(lline_col_min - rline_col_min); - const int dx_llol = std::abs(lline_col_min - lolline_col_min); - - if (dx_lr < delta_alignment) - left_right_aligned = true; - - if (dx_llol < delta_alignment) - left_lol_aligned = true; - - if (left_right_aligned && left_lol_aligned) - { - const int left_right_col_min = std::min(lline_col_min, rline_col_min); - const int dx_lrc = std::abs(left_right_col_min - cline_col_min); - const float l_char_width = 1.5f * lines(l).char_width(); - - if (dx_lrc > l_char_width && - dx_lrc < 3.0f * l_char_width && - cline_col_min > rline_col_min && - cline_col_min > lline_col_min) - { - output(right_nbh) = right_nbh; - continue; - } - } - } - -//----------------------------------------------------------------------------- -// ___________________________ -// |___________________________| // ___________________ // |___________________| End of the paragraph - Current line // ________________________ @@ -443,8 +531,11 @@ namespace scribo { const int dx_lrc = std::abs(lline_col_max - cline_col_max); const int l_char_width = lines(l).char_width(); + const int dx_indent = std::abs(std::max(lline_col_min, + rline_col_min) - cline_col_min); if (dx_lrc > l_char_width && + dx_indent < 4 * delta_alignment && cline_col_max < lline_col_max && cline_col_min < lline_col_min && (lline_col_min > lolline_col_min || lol_is_left)) @@ -456,48 +547,216 @@ namespace scribo } + //----------------------------------------------------------------------------- // ___________________________ // |___________________________| -// ___________________________ -// |___________________________| // ________________________ // |________________________| +// ___________________________ +// |___________________________| +// ___________________________ +// |___________________________| // // Simple case : paragraphs are justified on the left. We try to find any -// indentation like above at the end of a column. +// indentation like above. // //----------------------------------------------------------------------------- - if (left_nbh == l) { - const value::int_u16 ror_nbh = right(right_nbh); - const box2d& ror_line_bbox = lines(ror_nbh).bbox(); - const int rorline_col_min = ror_line_bbox.pmin().col(); + // Check if the current line neighbors are aligned + bool left_right_aligned = false; + bool left_lol_aligned = false; + const int dx_lr = std::abs(lline_col_min - rline_col_min); + const int dx_llol = std::abs(lline_col_min - lolline_col_min); - bool right_ror_min_aligned = false; - const int dx_rror_min = std::abs(rline_col_min - rorline_col_min); + if (dx_lr < delta_alignment) + left_right_aligned = true; - if (dx_rror_min < delta_alignment) - right_ror_min_aligned = true; + if (dx_llol < delta_alignment) + left_lol_aligned = true; - if (right_ror_min_aligned) + if (left_right_aligned && left_lol_aligned) { - const int right_ror_col_min = std::min(rline_col_min, rorline_col_min); - const int dx_rrorc = std::abs(right_ror_col_min - cline_col_min); + const int left_right_col_min = std::min(lline_col_min, rline_col_min); + const int dx_lrc = std::abs(left_right_col_min - cline_col_min); const float l_char_width = 1.5f * lines(l).char_width(); - if (dx_rrorc > l_char_width && - dx_rrorc < 3.0f * l_char_width && + if (dx_lrc > l_char_width && + !may_have_another_left_link(right, right_nbh, l, lines) && + dx_lrc < 10.0f * l_char_width && cline_col_min > rline_col_min && - cline_col_max >= rline_col_max) + cline_col_min > lline_col_min) { - output(right_nbh) = right_nbh; + const value::int_u16 out_right_nbh = output(right_nbh); + + if (out_right_nbh != l) + right(l) = l; + else + output(right_nbh) = right_nbh; continue; } } } - } + +//----------------------------------------------------------------------------- +// ___________________________ +// |___________________________| +// ___________________________ +// |___________________________| +// ________________________ +// |________________________| +// ___________________________ +// |___________________________| +// +// Simple case : paragraphs are justified on the left. We try to find any +// indentation like above. +// +//----------------------------------------------------------------------------- + + { + const value::int_u16 ror_nbh = right(right_nbh); + const box2d& ror_line_bbox = lines(ror_nbh).bbox(); + const int rorline_col_min = ror_line_bbox.pmin().col(); + + bool right_ror_min_aligned = false; + bool left_right_aligned = false; + const int dx_lr = std::abs(lline_col_min - rline_col_min); + const int dx_rror_min = std::abs(rline_col_min - rorline_col_min); + + if (dx_rror_min < delta_alignment) + right_ror_min_aligned = true; + + if (dx_lr < delta_alignment) + left_right_aligned = true; + + if (right_ror_min_aligned && left_right_aligned && + ror_nbh != right_nbh) + { + const int left_right_col_min = std::min(lline_col_min, rline_col_min); + const int dx_lrc = std::abs(left_right_col_min - cline_col_min); + const float l_char_width = 1.5f * lines(l).char_width(); + + if (dx_lrc > l_char_width && + !may_have_another_left_link(right, right_nbh, l, lines) && + dx_lrc < 10.0f * l_char_width && + cline_col_min > rline_col_min && + cline_col_min > lline_col_min) + { + const value::int_u16 out_right_nbh = output(right_nbh); + + if (out_right_nbh != l) + right(l) = l; + else + output(right_nbh) = right_nbh; + continue; + } + } + } + +//----------------------------------------------------------------------------- +// ___________________________ +// |___________________________| +// ___________ +// |___________| +// ________________________ +// |________________________| +// ___________________________ +// |___________________________| +// +// Simple case : paragraphs are justified on the left. We try to find any +// indentation like above. +// +//----------------------------------------------------------------------------- + + { + const value::int_u16 ror_nbh = right(right_nbh); + const box2d& ror_line_bbox = lines(ror_nbh).bbox(); + const int rorline_col_min = ror_line_bbox.pmin().col(); + + bool left_ror_aligned = false; + const int dx_lror = std::abs(lline_col_min - rorline_col_min); + + if (dx_lror < delta_alignment) + left_ror_aligned = true; + + if (left_ror_aligned) + { + const int left_ror_col_min = std::min(lline_col_min, rorline_col_min); + const int dx_lrorc = std::abs(left_ror_col_min - cline_col_min); + const float l_char_width = 1.5f * lines(l).char_width(); + const int dx_lrorr = std::abs(left_ror_col_min - rline_col_min); + const int dx_crmax = std::abs(rline_col_max - cline_col_max); + + if (dx_lrorc > l_char_width && + dx_lrorr > 5 * l_char_width && + dx_lrorr > dx_lrorc && + dx_crmax > 5 * l_char_width && + !may_have_another_left_link(right, right_nbh, l, lines) && + dx_lrorc < 10.0f * l_char_width && + cline_col_min > rorline_col_min && + cline_col_min > lline_col_min) + { + right(right_nbh) = right_nbh; + continue; + } + } + } + + +// Strange case + { + if (rline_col_min > current_line_bbox.pcenter().col() + && !may_have_another_left_link(right, right_nbh, l, lines) + && cline_col_max < rline_col_max + && output(right_nbh) == l) + { + output(right_nbh) = right_nbh; + } + } + +//----------------------------------------------------------------------------- +// ___________________________ +// |___________________________| +// ___________________________ +// |___________________________| +// ________________________ +// |________________________| +// +// Simple case : paragraphs are justified on the left. We try to find any +// indentation like above at the end of a column. +// +//----------------------------------------------------------------------------- + + if (left_nbh == l) + { + const value::int_u16 ror_nbh = right(right_nbh); + const box2d& ror_line_bbox = lines(ror_nbh).bbox(); + const int rorline_col_min = ror_line_bbox.pmin().col(); + + bool right_ror_min_aligned = false; + const int dx_rror_min = std::abs(rline_col_min - rorline_col_min); + + if (dx_rror_min < delta_alignment) + right_ror_min_aligned = true; + + if (right_ror_min_aligned) + { + const int right_ror_col_min = std::min(rline_col_min, rorline_col_min); + const int dx_rrorc = std::abs(right_ror_col_min - cline_col_min); + const float l_char_width = 1.5f * lines(l).char_width(); + + if (dx_rrorc > l_char_width && + dx_rrorc < 10.0f * l_char_width && + cline_col_min > rline_col_min && + cline_col_max >= rline_col_max) + { + output(right_nbh) = right_nbh; + continue; + } + } + } + } // Only debug @@ -505,8 +764,21 @@ namespace scribo { image2d<value::rgb8> debug = data::convert(value::rgb8(), input); + const util::array<value::int_u16> backup = output; for (unsigned i = 0; i < output.nelements(); ++i) - output(i) = internal::find_root(output, i); + { + const value::int_u16 current_neighbor = backup(i); + output(i) = internal::find_root(output, i); + const value::int_u16 root_index = output(i); + + for (unsigned j = 0; j < right.nelements(); ++j) + { + if (i != j && + current_neighbor != i && + right(j) == i) + internal::set_root(output, j, root_index); + } + } mln::util::array<accu::shape::bbox<point2d> > nbbox(output.nelements()); for (unsigned i = 0; i < nlines; ++i) @@ -690,7 +962,7 @@ namespace scribo int dmax = 1.5f * lines(i).x_height(); // Starting points in the current line box - point2d c = rbbox(i).pcenter(); + point2d c(rbbox(i).pcenter()); point2d q(rbbox(i).pmin().row() + ((c.row() - rbbox(i).pmin().row()) / 4), c.col()); int @@ -830,7 +1102,7 @@ namespace scribo int dmax = 1.5f * lines(i).x_height(); // Starting points in the current line box - point2d c = rbbox(i).pcenter(); + point2d c(rbbox(i).pcenter()); point2d q(rbbox(i).pmax().row() - ((rbbox(i).pmax().row() - c.row()) / 4), c.col()); int @@ -952,7 +1224,9 @@ namespace scribo inline void finalize_links(util::array<value::int_u16>& left, util::array<value::int_u16>& right, - const util::array< line_info<L> >& lines) + const util::array< line_info<L> >& lines// , + // const image2d<bool>& input + ) { const unsigned nlines = lines.nelements(); @@ -979,6 +1253,17 @@ namespace scribo v = i; } } + + // image2d<value::rgb8> left_links = data::convert(value::rgb8(), input); + // image2d<value::rgb8> right_links = data::convert(value::rgb8(), input); + + // for (unsigned l = 0; l < nlines; ++l) + // { + // mln::draw::line(left_links, lines(l).bbox().pcenter(), lines(left(l)).bbox().pcenter(), literal::red); + // mln::draw::line(right_links, lines(l).bbox().pcenter(), lines(right(l)).bbox().pcenter(), literal::red); + // } + // mln::io::ppm::save(left_links, "out_left_links.ppm"); + // mln::io::ppm::save(right_links, "out_right_links.ppm"); } template <typename L> @@ -1016,7 +1301,8 @@ namespace scribo std::cout << "Linking right" << std::endl; process_right_link(blocks, rbbox, lines_info, right); std::cout << "Finalizing links" << std::endl; - finalize_links(left, right, lines_info); + finalize_links(left, right, lines_info// , input + ); // std::cout << "Finalizing merging" << std::endl; // finalize_line_merging(left, right, lines); std::cout << "Extracting paragraphs" << std::endl; -- 1.5.6.5

13 years, 8 months

last-svn-commit-881-g973e5ac Improve paragraph grouping for historical documents.

by Julien Marquegnies

* scribo/core/line_info.hh, * scribo/text/merging.hh, * scribo/text/paragraphs.hh: Here. --- scribo/ChangeLog | 8 + scribo/scribo/core/line_info.hh | 18 ++- scribo/scribo/text/merging.hh | 16 +- scribo/scribo/text/paragraphs.hh | 388 ++++++++++++++++++++++++++++++++++---- 4 files changed, 378 insertions(+), 52 deletions(-) diff --git a/scribo/ChangeLog b/scribo/ChangeLog index a28bdf7..5e29749 100644 --- a/scribo/ChangeLog +++ b/scribo/ChangeLog @@ -1,3 +1,11 @@ +2011-05-18 Julien Marquegnies <marquegnies(a)lrde.epita.fr> + + Improve paragraph grouping for historical documents. + + * scribo/core/line_info.hh, + * scribo/text/merging.hh, + * scribo/text/paragraphs.hh: Here. + 2011-05-17 Guillaume Lazzara <z(a)lrde.epita.fr> Fix use of skeleton_constrained. diff --git a/scribo/scribo/core/line_info.hh b/scribo/scribo/core/line_info.hh index e91e1c3..9017174 100644 --- a/scribo/scribo/core/line_info.hh +++ b/scribo/scribo/core/line_info.hh @@ -1000,21 +1000,22 @@ namespace scribo const int dy = bb1.pmax().row() - bb2.pmax().row(); // The two characters must be distinct - if (space < 0) - return false; + // if (space < 0) + // return false; if (// Approximately the same width - ((std::max(w1, w2) / std::min(w1, w2)) > 1.1f || + ((std::max(w1, w2) / std::min(w1, w2)) > 1.3 || // One character must not be smaller than the space between // the two characters (w1 < space || w2 < space)) // If the two characters have a different width they must also // have a different height - && not (std::max(h1, h2) / std::min(h1, h2) <= 1.5f)) + && not (std::max(h1, h2) / std::min(h1, h2) <= 1.7f)) return false; // Approximately aligned on baseline - if (std::abs(dy) > 10) + if (std::abs(dy) > 10 && + not (std::max(h1, h2) / std::min(h1, h2) <= 1.7f)) return false; return true; @@ -1049,6 +1050,10 @@ namespace scribo float min_base = 0.0f; const unsigned clusters_b_nelements = clusters_b.nelements(); + if (clusters_b_nelements >= 3) + return data_->baseline_clusters_.mean(); + + for (unsigned i = 0; i < clusters_b_nelements; ++i) { const unsigned clusters_b_i_nelements = clusters_b[i].nelements(); @@ -1086,6 +1091,9 @@ namespace scribo float max_mean = 0.0f; const unsigned clusters_m_nelements = clusters_m.nelements(); + if (clusters_m_nelements >= 3) + return data_->meanline_clusters_.mean(); + for (unsigned i = 0; i < clusters_m_nelements; ++i) { const unsigned clusters_m_i_nelements = clusters_m[i].nelements(); diff --git a/scribo/scribo/text/merging.hh b/scribo/scribo/text/merging.hh index f691188..2763882 100644 --- a/scribo/scribo/text/merging.hh +++ b/scribo/scribo/text/merging.hh @@ -461,8 +461,8 @@ namespace scribo if (l_cur_height < l_ted_x_height && l_cur_height > 0.05f * l_ted_x_height && float(l_cur_width) / float(l_cur.card()) < l_ted.char_width() - && dx < l_ted_cw - && l_cur_pmin.row() < l_ted_pmax.row()) + && dx < 2 * l_ted_cw + && l_cur_pmin.row() < l_ted.baseline()) { l_cur.update_type(line::Punctuation); return true; @@ -741,11 +741,12 @@ namespace scribo // vertically aligned // Obviously no separators between the two lines if ((l_info.card() <= 5 || - (std::abs(l_info.baseline() - mc_info.baseline()) < 5 - && std::abs(l_info.meanline() - mc_info.meanline()) < 5)) - && dx < l_ted_cw && dy < 0 - && not (l_info.holder().components().has_separators() - && between_separators(l_info, mc_info))) + (std::abs(l_info.baseline() - mc_info.baseline()) + < 5 && std::abs(l_info.meanline() - + mc_info.meanline()) < 5)) + && dx < l_ted_cw && dy < 0 + && not (l_info.holder().components().has_separators() + && between_separators(l_info, mc_info))) l = do_union(lines, l, mc, parent); // } @@ -1047,7 +1048,6 @@ namespace scribo // ts = t.stop(); // std::cout << "time " << ts << std::endl; - lines.force_stats_update(); return lines; diff --git a/scribo/scribo/text/paragraphs.hh b/scribo/scribo/text/paragraphs.hh index 6c9285b..9a59e02 100644 --- a/scribo/scribo/text/paragraphs.hh +++ b/scribo/scribo/text/paragraphs.hh @@ -30,30 +30,140 @@ namespace scribo //------------------------------------- // Extracting root of links //------------------------------------- - template <typename T> - inline - unsigned - find_root(util::array<T>& parent, unsigned x) - { - unsigned tmp_x = x; + template <typename T> + inline + unsigned + find_root(util::array<T>& parent, unsigned x) + { + unsigned tmp_x = x; - while (parent(tmp_x) != tmp_x) - tmp_x = parent(tmp_x); + while (parent(tmp_x) != tmp_x) + tmp_x = parent(tmp_x); - while (parent(x) != x) - { - const unsigned tmp = parent(x); - x = parent(x); - parent(tmp) = tmp_x; - } + while (parent(x) != x) + { + const unsigned tmp = parent(x); + x = parent(x); + parent(tmp) = tmp_x; + } + + return x; + } - return x; + template <typename T> + inline + void + set_root(util::array<T>& parent, unsigned x, const unsigned root) + { + while (parent(x) != x && parent(x) != root) + { + const unsigned tmp = parent(x); + x = parent(x); + parent(tmp) = root; } + + parent(x) = root; + } + } namespace filter { + template <typename L> + inline + bool + between_horizontal_separator(const scribo::line_info<L>& l1, + const scribo::line_info<L>& l2) + { + // No separators found in image. + mln_precondition(l1.holder().components().has_separators()); + + const box2d& l1_bbox = l1.bbox(); + const box2d& l2_bbox = l2.bbox(); + + unsigned + row1 = l1_bbox.pcenter().row(), + row2 = l2_bbox.pcenter().row(); + const mln_ch_value(L, bool)& + separators = l1.holder().components().separators(); + + unsigned row; + unsigned col_ptr; + unsigned left_col_ptr; + unsigned right_col_ptr; + unsigned end; + + if (row1 < row2) + { + row1 = l1_bbox.pmax().row(); + row2 = l2_bbox.pmin().row(); + + const unsigned quarter = + ((l1_bbox.pcenter().col() - l1_bbox.pmin().col()) >> 2); + + row = l1_bbox.pcenter().row(); + col_ptr = l1_bbox.pcenter().col(); + left_col_ptr = l1_bbox.pmin().col() + quarter; + right_col_ptr = l1_bbox.pmax().col() - quarter; + end = row2; + } + else + { + row2 = l2_bbox.pmax().row(); + row1 = l1_bbox.pmin().row(); + + const unsigned quarter = + ((l2_bbox.pcenter().col() - l2_bbox.pmin().col()) >> 2); + + row = l2_bbox.pcenter().row(); + col_ptr = l2_bbox.pcenter().col(); + left_col_ptr = l2_bbox.pmin().col() + quarter; + right_col_ptr = l2_bbox.pmax().col() - quarter; + end = row1; + } + + // If sep_ptr is true, then a separator is reached. + while (row < end) + { + ++row; + if (separators.at_(row, col_ptr) + || separators.at_(row, left_col_ptr) + || separators.at_(row, right_col_ptr)) + return true; + } + + return false; + } + + + template <typename L> + bool may_have_another_left_link(const line_links<L>& right, + const line_id_t& index, + const line_id_t& current_line, + const line_set<L>& lines) + { + const line_info<L>& l = lines(current_line); + const point2d& pmin = l.bbox().pmin(); + const unsigned x1 = l.x_height(); + + for_all_links(i, right) + if (i != index && right(i) == index) + { + const line_info<L>& l_info = lines(i); + const unsigned x2 = l_info.x_height(); + + const float delta_max = 0.5f * std::min(x1, x2); + + if (l_info.bbox().pmin().col() < pmin.col() + && std::abs(l.baseline() - l_info.baseline()) < delta_max + ) + return true; + } + + return false; + } + //--------------------------------------------------------------------- // This method aims to cut the links between lines that do not fit the // different criteria @@ -62,7 +172,7 @@ namespace scribo template <typename L> inline void paragraph_links(const line_links<L>& left, - const line_links<L>& right, + line_links<L>& right, line_links<L>& output, const line_set<L>& lines) { @@ -83,9 +193,27 @@ namespace scribo { // Neighbors - const line_id_t left_nbh = output(l); - const line_id_t right_nbh = right(l); - const line_id_t lol_nbh = output(left_nbh); + line_id_t left_nbh = output(l); + line_id_t right_nbh = right(l); + line_id_t lol_nbh = output(left_nbh); + + const line_info<L>& left_line = lines(left_nbh); + const line_info<L>& current_line = lines(l); + const line_info<L>& right_line = lines(right_nbh); + + if (right_line.holder().components().has_separators() && + between_horizontal_separator(right_line, current_line)) + { + output(right_nbh) = right_nbh; + right_nbh = l; + } + if (current_line.holder().components().has_separators() && + between_horizontal_separator(current_line, left_line)) + { + output(l) = l; + left_nbh = l; + lol_nbh = l; + } // Line features const float x_height = lines(l).x_height(); @@ -112,6 +240,7 @@ namespace scribo // Maximal x variation to consider two lines vertically aligned const int delta_alignment = cline_cw; + // Checks the baseline distances of the two neighbors { // Current line baseline @@ -151,7 +280,7 @@ namespace scribo // and its right neighbor if (right_distance > 1.4f * ror_distance && std::max(ror_x_height, right_x_height) < - 1.2f * std::min(ror_x_height, right_x_height) + 1.4f * std::min(ror_x_height, right_x_height) && output(right_nbh) == l) { output(right_nbh) = right_nbh; @@ -184,7 +313,7 @@ namespace scribo // Condition to cut the link between the current line and // its right neighbor - if ((max_x_height > min_x_height * 1.2f) && + if ((max_x_height > min_x_height * 1.4f) && !(max_char_width <= 1.2f * min_char_width)) { if (output(right_nbh) == l) @@ -220,7 +349,7 @@ namespace scribo // and its left neighbor if (left_distance > 1.4f * lol_distance && std::max(lol_x_height, left_x_height) < - 1.2f * std::min(lol_x_height, left_x_height)) + 1.4f * std::min(lol_x_height, left_x_height)) { output(l) = l; continue; @@ -252,7 +381,7 @@ namespace scribo // Condition to cut the link between the current line and // its left neighbor - if ((max_x_height > min_x_height * 1.2f) && + if ((max_x_height > min_x_height * 1.4f) && !(max_char_width <= 1.2f * min_char_width)) { output(l) = l; @@ -264,18 +393,18 @@ namespace scribo continue; } // The current line has at least one left and one right neighbor - else // if (delta_baseline_max >= delta_baseline_min) + else // if (delta_baseline_max >= 1.1 * delta_baseline_min) { // Distance between the left and the current line - const float left_distance = - lines(left_nbh).meanline() - lines(l).baseline(); + const float + left_distance = left_line_bbox.pcenter().row() - current_line_bbox.pcenter().row(); // Distance between the right and the current line - const float right_distance = - lines(l).meanline() - lines(right_nbh).baseline(); + const float + right_distance = current_line_bbox.pcenter().row() - right_line_bbox.pcenter().row();; // If the left line is too far compared to the right one // we cut the link with it - if (left_distance > 1.2f * right_distance + if (left_distance > 1.5f * right_distance && std::max(x_height, left_x_height) > 1.2f * std::min(x_height, left_x_height)) { output(l) = l; @@ -283,8 +412,8 @@ namespace scribo } // If the right line is too far compared to the left one // we cut the link with it - else if (right_distance > 1.2f * left_distance - && std::max(x_height, right_x_height) > 1.2f * std::min(x_height, right_x_height) + else if (right_distance > 1.5f * left_distance + && std::max(x_height, right_x_height) >= 1.2f * std::min(x_height, right_x_height) && output(right_nbh) == l) { output(right_nbh) = right_nbh; @@ -303,7 +432,7 @@ namespace scribo const float min_x_height = std::min(x_height, left_x_height); const float max_x_height = std::max(x_height, left_x_height); - if ((max_x_height > min_x_height * 1.2f) && + if ((max_x_height > min_x_height * 1.4f) && !(cw_max <= 1.2f * cw_min)) { output(l) = l; @@ -316,7 +445,7 @@ namespace scribo const float cw_max = std::max(rline_cw, cline_cw); const float cw_min = std::min(rline_cw, cline_cw); - if ((max_x_height > min_x_height * 1.2f) + if ((max_x_height > min_x_height * 1.4f) && !(cw_max <= 1.2f * cw_min) && output(right_nbh) == l) { @@ -332,7 +461,7 @@ namespace scribo const float min_x_height = std::min(x_height, right_x_height); const float max_x_height = std::max(x_height, right_x_height); - if ((max_x_height > min_x_height * 1.2f) + if ((max_x_height > min_x_height * 1.4f) && !(cw_max <= 1.2f * cw_min) && output(right_nbh) == l) { @@ -346,7 +475,7 @@ namespace scribo const float cw_max = std::max(lline_cw, cline_cw); const float cw_min = std::min(lline_cw, cline_cw); - if ((max_x_height > min_x_height * 1.2f) + if ((max_x_height > min_x_height * 1.4f) && !(cw_max <= 1.2f * cw_min)) { output(l) = l; @@ -445,8 +574,11 @@ namespace scribo { const int dx_lrc = std::abs(lline_col_max - cline_col_max); const int l_char_width = lines(l).char_width(); + const int dx_indent = std::abs(std::max(lline_col_min, + rline_col_min) - cline_col_min); if (dx_lrc > l_char_width && + dx_indent < 4 * delta_alignment && cline_col_max < lline_col_max && cline_col_min < lline_col_min && (lline_col_min > lolline_col_min || lol_is_left)) @@ -457,6 +589,172 @@ namespace scribo } } +//----------------------------------------------------------------------------- +// ___________________________ +// |___________________________| +// ________________________ +// |________________________| +// ___________________________ +// |___________________________| +// ___________________________ +// |___________________________| +// +// Simple case : paragraphs are justified on the left. We try to find any +// indentation like above. +// +//----------------------------------------------------------------------------- + + { + // Check if the current line neighbors are aligned + bool left_right_aligned = false; + bool left_lol_aligned = false; + const int dx_lr = std::abs(lline_col_min - rline_col_min); + const int dx_llol = std::abs(lline_col_min - lolline_col_min); + + if (dx_lr < delta_alignment) + left_right_aligned = true; + + if (dx_llol < delta_alignment) + left_lol_aligned = true; + + if (left_right_aligned && left_lol_aligned) + { + const int left_right_col_min = std::min(lline_col_min, rline_col_min); + const int dx_lrc = std::abs(left_right_col_min - cline_col_min); + const float l_char_width = 1.5f * lines(l).char_width(); + + if (dx_lrc > l_char_width && + !may_have_another_left_link(right, right_nbh, l, lines) && + dx_lrc < 10.0f * l_char_width && + cline_col_min > rline_col_min && + cline_col_min > lline_col_min) + { + const line_id_t out_right_nbh = output(right_nbh); + + if (out_right_nbh != l) + right(l) = l; + else + output(right_nbh) = right_nbh; + continue; + } + } + } + +//----------------------------------------------------------------------------- +// ___________________________ +// |___________________________| +// ___________________________ +// |___________________________| +// ________________________ +// |________________________| +// ___________________________ +// |___________________________| +// +// Simple case : paragraphs are justified on the left. We try to find any +// indentation like above. +// +//----------------------------------------------------------------------------- + + { + const line_id_t ror_nbh = right(right_nbh); + const box2d& ror_line_bbox = lines(ror_nbh).bbox(); + const int rorline_col_min = ror_line_bbox.pmin().col(); + + bool right_ror_min_aligned = false; + bool left_right_aligned = false; + const int dx_lr = std::abs(lline_col_min - rline_col_min); + const int dx_rror_min = std::abs(rline_col_min - rorline_col_min); + + if (dx_rror_min < delta_alignment) + right_ror_min_aligned = true; + + if (dx_lr < delta_alignment) + left_right_aligned = true; + + if (right_ror_min_aligned && left_right_aligned && + ror_nbh != right_nbh) + { + const int left_right_col_min = std::min(lline_col_min, rline_col_min); + const int dx_lrc = std::abs(left_right_col_min - cline_col_min); + const float l_char_width = 1.5f * lines(l).char_width(); + + if (dx_lrc > l_char_width && + !may_have_another_left_link(right, right_nbh, l, lines) && + dx_lrc < 10.0f * l_char_width && + cline_col_min > rline_col_min && + cline_col_min > lline_col_min) + { + const line_id_t out_right_nbh = output(right_nbh); + + if (out_right_nbh != l) + right(l) = l; + else + output(right_nbh) = right_nbh; + continue; + } + } + } + +//----------------------------------------------------------------------------- +// ___________________________ +// |___________________________| +// ___________ +// |___________| +// ________________________ +// |________________________| +// ___________________________ +// |___________________________| +// +// Simple case : paragraphs are justified on the left. We try to find any +// indentation like above. +// +//----------------------------------------------------------------------------- + + { + const line_id_t ror_nbh = right(right_nbh); + const box2d& ror_line_bbox = lines(ror_nbh).bbox(); + const int rorline_col_min = ror_line_bbox.pmin().col(); + + bool left_ror_aligned = false; + const int dx_lror = std::abs(lline_col_min - rorline_col_min); + + if (dx_lror < delta_alignment) + left_ror_aligned = true; + + if (left_ror_aligned) + { + const int left_ror_col_min = std::min(lline_col_min, rorline_col_min); + const int dx_lrorc = std::abs(left_ror_col_min - cline_col_min); + const float l_char_width = 1.5f * lines(l).char_width(); + const int dx_lrorr = std::abs(left_ror_col_min - rline_col_min); + const int dx_crmax = std::abs(rline_col_max - cline_col_max); + + if (dx_lrorc > l_char_width && + dx_lrorr > 5 * l_char_width && + dx_lrorr > dx_lrorc && + dx_crmax > 5 * l_char_width && + !may_have_another_left_link(right, right_nbh, l, lines) && + dx_lrorc < 10.0f * l_char_width && + cline_col_min > rorline_col_min && + cline_col_min > lline_col_min) + { + right(right_nbh) = right_nbh; + continue; + } + } + } + + +// Strange case + { + if (rline_col_min > current_line_bbox.pcenter().col() + && !may_have_another_left_link(right, right_nbh, l, lines) + && cline_col_max < rline_col_max + && output(right_nbh) == l) + { + output(right_nbh) = right_nbh; + } + } //----------------------------------------------------------------------------- // ___________________________ @@ -490,7 +788,7 @@ namespace scribo const float l_char_width = 1.5f * lines(l).char_width(); if (dx_rrorc > l_char_width && - dx_rrorc < 3.0f * l_char_width && + dx_rrorc < 10.0f * l_char_width && cline_col_min > rline_col_min && cline_col_max >= rline_col_max) { @@ -501,14 +799,26 @@ namespace scribo } } - // Only debug // { // image2d<value::rgb8> debug = data::convert(value::rgb8(), input); - // for (unsigned i = 0; i < output.nelements(); ++i) - // output(i) = scribo::make::internal::find_root(output, i); + // const util::array<value::int_u16> backup = output; + // for (unsigned i = 0; i < output.nelements(); ++i) + // { + // const value::int_u16 current_neighbor = backup(i); + // output(i) = internal::find_root(output, i); + // const value::int_u16 root_index = output(i); + + // for (unsigned j = 0; j < right.nelements(); ++j) + // { + // if (i != j && + // current_neighbor != i && + // right(j) == i) + // internal::set_root(output, j, root_index); + // } + // } // mln::util::array<accu::shape::bbox<point2d> > nbbox(output.nelements()); // for_all_lines(l, lines) -- 1.5.6.5

13 years, 8 months

branch icdar/hdlac2011 created: last-svn-commit-880-g853ed71

by lazzara＠lrde.epita.fr

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "Olena, a generic and efficient image processing platform". The branch icdar/hdlac2011 has been created at 853ed71516fef6f26d0ba5cd80b35a74f6c53269 (commit) - Log ----------------------------------------------------------------- 853ed71 Add new sample programs. e94251a Fix use of skeleton_constrained. 23a40a7 Add new tests. cc99fdd New routine to extract both thin and thick separators. 7f93eff scribo/primitive/extract/non_text_hdoc.hh: New routine for historical documents. 9562a2a New object filter. cb19d2d Various small fixes in Scribo. 129baa2 Add holder information to component_info. 316f515 New specific toolchain for historical documents. ----------------------------------------------------------------------- hooks/post-receive -- Olena, a generic and efficient image processing platform

13 years, 8 months

branch unstable/scribo updated: last-svn-commit-873-g3e49926

by lazzara＠lrde.epita.fr

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "Olena, a generic and efficient image processing platform". The branch unstable/scribo has been updated via 3e4992613401cb2f4332d159cd29b1655074f997 (commit) via 23ddba54a398be77147c97cc759bc94d684b4134 (commit) via f18122560baaac00ecf531088d2874e6a77d4b79 (commit) via 2ec107afdf1e4db1ca1e06b9684695a1076b4b28 (commit) via 6790b23139b526e7318b555459a1d5d86775e6b2 (commit) via c287ea3f46551cd53d27868439a609b526223610 (commit) from 9d8ad622a97eb18dec9decd8461b3b05c5f4566c (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- 3e49926 Make code for icdar 2009 compile. 23ddba5 Add a new tool. f181225 Fix precision issue with algebra::vec conversion. 2ec107a Fix use of skeleton_constrained. 6790b23 Add fastest versions for skeleton constrained related algorithms. c287ea3 Update support for PAGE XML format. ----------------------------------------------------------------------- Summary of changes: milena/ChangeLog | 21 ++ milena/mln/core/point.hh | 32 +-- milena/mln/morpho/skeleton_constrained.hh | 271 ++++++++++++++++---- milena/mln/topo/skeleton/crest.hh | 230 ++++++++++++++--- milena/mln/topo/skeleton/is_simple_point.hh | 221 +++++++---------- milena/sandbox/ChangeLog | 9 + milena/sandbox/icdar/2009/hsc/clean_input.hh | 2 +- milena/sandbox/icdar/2009/hsc/ws_to_wsl.hh | 14 +- milena/tests/geom/rotate.cc | 22 +- milena/tests/topo/skeleton/is_simple_point.cc | 20 +- scribo/ChangeLog | 31 +++ scribo/scribo/core/tag/line.hh | 13 +- scribo/scribo/estim/components_features.hh | 4 +- scribo/scribo/estim/internal/compute_skeleton.hh | 4 +- .../scribo/io/xml/internal/compute_text_colour.hh | 32 ++-- .../io/xml/internal/extended_page_xml_visitor.hh | 4 +- scribo/scribo/io/xml/internal/full_xml_visitor.hh | 4 +- scribo/scribo/io/xml/internal/page_xml_visitor.hh | 77 ++++-- scribo/scribo/io/xml/internal/print_box_coords.hh | 12 +- .../scribo/io/xml/internal/print_page_preambule.hh | 42 ++-- scribo/src/misc/Makefile.am | 6 +- .../src/misc/closure_rectangle.cc | 28 ++- 22 files changed, 743 insertions(+), 356 deletions(-) copy milena/tests/io/tiff/tiff2pbm.cc => scribo/src/misc/closure_rectangle.cc (74%) hooks/post-receive -- Olena, a generic and efficient image processing platform

13 years, 8 months

last-svn-commit-873-g3e49926 Make code for icdar 2009 compile.

by Guillaume Lazzara

* icdar/2009/hsc/clean_input.hh: Use pcenter(). * icdar/2009/hsc/ws_to_wsl.hh: Update call to make::p_vertices_with_mass_centers. --- milena/sandbox/ChangeLog | 9 +++++++++ milena/sandbox/icdar/2009/hsc/clean_input.hh | 2 +- milena/sandbox/icdar/2009/hsc/ws_to_wsl.hh | 14 +++++++------- 3 files changed, 17 insertions(+), 8 deletions(-) diff --git a/milena/sandbox/ChangeLog b/milena/sandbox/ChangeLog index 34e60e0..48984a8 100644 --- a/milena/sandbox/ChangeLog +++ b/milena/sandbox/ChangeLog @@ -1,3 +1,12 @@ +2011-05-18 Guillaume Lazzara <z(a)lrde.epita.fr> + + Make code for icdar 2009 compile. + + * icdar/2009/hsc/clean_input.hh: Use pcenter(). + + * icdar/2009/hsc/ws_to_wsl.hh: Update call to + make::p_vertices_with_mass_centers. + 2010-06-21 Yann Jacquelet <jacquelet(a)lrde.epita.fr> Add milena library components. diff --git a/milena/sandbox/icdar/2009/hsc/clean_input.hh b/milena/sandbox/icdar/2009/hsc/clean_input.hh index 1c5b572..c243a14 100644 --- a/milena/sandbox/icdar/2009/hsc/clean_input.hh +++ b/milena/sandbox/icdar/2009/hsc/clean_input.hh @@ -34,7 +34,7 @@ namespace mln ok(l) = true; // into bg continue; } - int center_col = box[l].center().col(); + int center_col = box[l].pcenter().col(); if (center_col >= max_col - 3 || center_col <= 3) ok(l) = true; diff --git a/milena/sandbox/icdar/2009/hsc/ws_to_wsl.hh b/milena/sandbox/icdar/2009/hsc/ws_to_wsl.hh index c488f9f..35b8675 100644 --- a/milena/sandbox/icdar/2009/hsc/ws_to_wsl.hh +++ b/milena/sandbox/icdar/2009/hsc/ws_to_wsl.hh @@ -136,7 +136,7 @@ namespace mln def::coord col() const { mln_invariant(valid_); - return static_cast<def::coord>(0.49999f + mean_[1]); // col = coord #1 + return static_cast<def::coord>(0.49999f + mean_[1]); // col = coord #1 } /// Get the covariance matrix. @@ -184,7 +184,7 @@ namespace mln algebra::mat<2,2,float> cov_, cov_1_; float lambda1_, lambda2_; }; - + template <typename L> @@ -264,7 +264,7 @@ namespace mln typedef util::graph G; G gr = make::region_adjacency_graph(ws, c8(), n_basins); - + util::array<mahalanobis> m; m = mahalanobis_from_images(ws, small, n_basins); @@ -311,7 +311,7 @@ namespace mln literal::red); typedef p_vertices< G, fun::i2v::array<point2d> > pv_t; - pv_t pv = make::p_vertices_with_mass_centers(ws, n_basins, gr); + pv_t pv = make::p_vertices_with_mass_centers(ws, gr); debug::draw_graph(cool, pv, literal::green, @@ -321,7 +321,7 @@ namespace mln } #endif // LOG - + util::array<unsigned> v_left (n_basins + 1), v_right(n_basins + 1); @@ -373,7 +373,7 @@ namespace mln } - + util::array<L> parent(n_basins + 1); // Computing parent. @@ -433,7 +433,7 @@ namespace mln unsigned s = input.nrows() / small.nrows(); image2d<L> ws_nol = morpho::elementary::dilation(ws, c8()); - + mln_piter(box2d) p(input.domain()); for_all(p) if (input(p)) -- 1.5.6.5

13 years, 8 months

last-svn-commit-872-g23ddba5 Add a new tool.

by Guillaume Lazzara

* src/misc/Makefile.am: Add new target. * src/misc/closure_rectanle.cc: New. --- scribo/ChangeLog | 8 +++++ scribo/src/misc/Makefile.am | 6 +++- .../src/misc/closure_rectangle.cc | 28 ++++++++++++------- 3 files changed, 30 insertions(+), 12 deletions(-) copy milena/tests/io/tiff/tiff2pbm.cc => scribo/src/misc/closure_rectangle.cc (74%) diff --git a/scribo/ChangeLog b/scribo/ChangeLog index 0b6042c..0412e63 100644 --- a/scribo/ChangeLog +++ b/scribo/ChangeLog @@ -1,5 +1,13 @@ 2011-05-17 Guillaume Lazzara <z(a)lrde.epita.fr> + Add a new tool. + + * src/misc/Makefile.am: Add new target. + + * src/misc/closure_rectanle.cc: New. + +2011-05-17 Guillaume Lazzara <z(a)lrde.epita.fr> + Fix use of skeleton_constrained. * scribo/estim/components_features.hh, diff --git a/scribo/src/misc/Makefile.am b/scribo/src/misc/Makefile.am index 0323aaa..e44660f 100644 --- a/scribo/src/misc/Makefile.am +++ b/scribo/src/misc/Makefile.am @@ -1,5 +1,5 @@ -# Copyright (C) 2009, 2010 EPITA Research and Development Laboratory -# (LRDE). +# Copyright (C) 2009, 2010, 2011 EPITA Research and Development +# Laboratory (LRDE). # # This file is part of Olena. # @@ -18,12 +18,14 @@ include $(top_srcdir)/scribo/scribo.mk noinst_PROGRAMS = \ + closure_rectangle \ dmap \ morpho \ negate \ rectangularity \ superpose +closure_rectangle_SOURCES = closure_rectangle.cc dmap_SOURCES = dmap.cc morpho_SOURCES = morpho.cc negate_SOURCES = negate.cc diff --git a/milena/tests/io/tiff/tiff2pbm.cc b/scribo/src/misc/closure_rectangle.cc similarity index 74% copy from milena/tests/io/tiff/tiff2pbm.cc copy to scribo/src/misc/closure_rectangle.cc index a7484dc..b25206f 100644 --- a/milena/tests/io/tiff/tiff2pbm.cc +++ b/scribo/src/misc/closure_rectangle.cc @@ -24,25 +24,33 @@ // executable file might be covered by the GNU General Public License. #include <mln/core/image/image2d.hh> -#include <mln/io/tiff/load.hh> -#include <mln/io/pbm/save.hh> -// #include <mln/data/convert.hh> +#include <mln/core/alias/neighb2d.hh> + +#include <mln/io/pbm/all.hh> + +#include <mln/win/rectangle2d.hh> +#include <mln/morpho/closing/structural.hh> + +#include <scribo/make/debug_filename.hh> int main(int argc, char *argv[]) { using namespace mln; - if (argc < 3) + if (argc != 4) { - std::cout << "Usage: " << argv[0] << " <input.tiff> <output.pbm>" << std::endl; + std::cout << argv[0] << " in.pbm out.pbm wsize" << std::endl; return 1; } - image2d<bool> ima; - io::tiff::load(ima, argv[1]); - io::pbm::save(ima, argv[2]); + unsigned wsize = atoi(argv[3]); + typedef image2d<bool> I; + + I in; + io::pbm::load(in, argv[1]); + + I out = morpho::closing::structural(in, win::rectangle2d(wsize, wsize)); -// io::ppm::save(ima, argv[2]); -// io::pbm::save(data::convert(bool(), ima), argv[2]); + io::pbm::save(out, argv[2]); } -- 1.5.6.5

13 years, 8 months

last-svn-commit-880-g853ed71 Add new sample programs.

by Guillaume Lazzara

* src/debug/Makefile.am: New targets. * src/debug/show_groups_bboxes_filtered.cc, * src/debug/show_text_lines_wo_seps.cc: New. --- milena/ChangeLog | 9 ++++ scribo/src/debug/Makefile.am | 16 ++++++ ...ps_bboxes.cc => show_groups_bboxes_filtered.cc} | 38 +++++++++++---- ...er_characters.cc => show_text_lines_wo_seps.cc} | 51 ++++++++++--------- 4 files changed, 80 insertions(+), 34 deletions(-) copy scribo/src/debug/{show_groups_bboxes.cc => show_groups_bboxes_filtered.cc} (75%) copy scribo/src/debug/{show_info_median_inter_characters.cc => show_text_lines_wo_seps.cc} (67%) diff --git a/milena/ChangeLog b/milena/ChangeLog index 5523182..e8430cd 100644 --- a/milena/ChangeLog +++ b/milena/ChangeLog @@ -1,5 +1,14 @@ 2011-05-17 Guillaume Lazzara <z(a)lrde.epita.fr> + Add new sample programs. + + * src/debug/Makefile.am: New targets. + + * src/debug/show_groups_bboxes_filtered.cc, + * src/debug/show_text_lines_wo_seps.cc: New. + +2011-05-17 Guillaume Lazzara <z(a)lrde.epita.fr> + Fix precision issue with algebra::vec conversion. * mln/core/point.hh: Round float values. diff --git a/scribo/src/debug/Makefile.am b/scribo/src/debug/Makefile.am index be8bb78..35beaf2 100644 --- a/scribo/src/debug/Makefile.am +++ b/scribo/src/debug/Makefile.am @@ -20,6 +20,7 @@ include $(top_srcdir)/scribo/scribo.mk noinst_PROGRAMS = \ show_components_bboxes \ show_groups_bboxes \ + show_groups_bboxes_filtered \ show_info_x_height \ show_info_median_inter_characters \ show_lines_boldness \ @@ -53,6 +54,7 @@ noinst_PROGRAMS = \ show_components_bboxes_SOURCES = show_components_bboxes.cc show_groups_bboxes_SOURCES = show_groups_bboxes.cc +show_groups_bboxes_filtered_SOURCES = show_groups_bboxes_filtered.cc show_info_x_height_SOURCES = show_info_x_height.cc show_info_median_inter_characters_SOURCES = show_info_median_inter_characters.cc show_lines_boldness_SOURCES = show_lines_boldness.cc @@ -127,6 +129,20 @@ if HAVE_QT show_text_lines_LDADD = $(LDADD) \ $(QT_LIBS) + noinst_PROGRAMS += show_text_lines_wo_seps + show_text_lines_wo_seps_SOURCES = show_text_lines_wo_seps.cc + show_text_lines_wo_seps_CPPFLAGS = $(AM_CPPFLAGS) \ + $(MAGICKXX_CPPFLAGS) \ + $(QT_CPPFLAGS) -DHAVE_QT + show_text_lines_wo_seps_CXXFLAGS = $(AM_CXXFLAGS) \ + $(QT_CXXFLAGS) + show_text_lines_wo_seps_LDFLAGS = $(AM_LDFLAGS) \ + $(MAGICKXX_LDFLAGS) \ + $(QT_LDFLAGS) + show_text_lines_wo_seps_LDADD = $(LDADD) \ + $(QT_LIBS) + + noinst_PROGRAMS += show_stoppers show_stoppers_SOURCES = show_stoppers.cc diff --git a/scribo/src/debug/show_groups_bboxes.cc b/scribo/src/debug/show_groups_bboxes_filtered.cc similarity index 75% copy from scribo/src/debug/show_groups_bboxes.cc copy to scribo/src/debug/show_groups_bboxes_filtered.cc index c7457ad..93b0c84 100644 --- a/scribo/src/debug/show_groups_bboxes.cc +++ b/scribo/src/debug/show_groups_bboxes_filtered.cc @@ -31,6 +31,8 @@ #include <scribo/primitive/link/internal/dmax_width_and_height.hh> #include <scribo/filter/object_links_bbox_h_ratio.hh> +#include <scribo/filter/objects_v_thick.hh> +#include <scribo/filter/objects_small.hh> int main(int argc, char *argv[]) { @@ -38,10 +40,11 @@ int main(int argc, char *argv[]) using namespace scribo; using namespace scribo::primitive; - if (argc != 3 && argc != 4) + if (argc != 3 && argc != 4 && argc != 5 && argc != 6) { - std::cerr << "Usage : " << argv[0] << " input.pbm out.pbm [hratio = 2.5]" << std::endl; - std::cerr << "Group components and makes sure size ratio is correct." << std::endl; + std::cerr << "Usage : " << argv[0] << " input.pbm out.pbm [hratio = 2.5] [min_comp_area = 3] [max_comp_height = 300]" << std::endl; + std::cerr << "Remove small and large components. Group components and makes sure " + << " size ratio is correct." << std::endl; return 1; } @@ -56,6 +59,18 @@ int main(int argc, char *argv[]) components = scribo::primitive::extract::components(input, c8(), ncomponents); + int + min_comp_area = 3, + max_comp_height = 300; + + if (argc >= 5) + min_comp_area = atoi(argv[4]); + if (argc >= 6) + max_comp_height = atoi(argv[5]); + + components = scribo::filter::components_small(components, min_comp_area); + components = scribo::filter::objects_v_thick(components, max_comp_height); + object_links<L> left_link = primitive::link::with_single_left_link_dmax_ratio( @@ -80,6 +95,7 @@ int main(int argc, char *argv[]) if (argc == 4) hratio = atof(argv[3]); + object_links<L> hratio_filtered_links = filter::object_links_bbox_h_ratio(merged_links, hratio); @@ -89,13 +105,15 @@ int main(int argc, char *argv[]) line_set<L> lines(groups); - image2d<bool> output; - initialize(output, input); - data::fill(output, false); + { + image2d<bool> output; + initialize(output, input); + data::fill(output, false); - for_all_lines(l, lines) - if (lines(l).is_valid()) - mln::draw::box(output, lines(l).bbox(), true); + for_all_lines(l, lines) + if (lines(l).is_valid()) + mln::draw::box(output, lines(l).bbox(), true); - io::pbm::save(output, argv[2]); + io::pbm::save(output, argv[2]); + } } diff --git a/scribo/src/debug/show_info_median_inter_characters.cc b/scribo/src/debug/show_text_lines_wo_seps.cc similarity index 67% copy from scribo/src/debug/show_info_median_inter_characters.cc copy to scribo/src/debug/show_text_lines_wo_seps.cc index 2ba240e..98ef05e 100644 --- a/scribo/src/debug/show_info_median_inter_characters.cc +++ b/scribo/src/debug/show_text_lines_wo_seps.cc @@ -1,5 +1,4 @@ -// Copyright (C) 2009, 2010, 2011 EPITA Research and Development -// Laboratory (LRDE) +// Copyright (C) 2011 EPITA Research and Development Laboratory (LRDE) // // This file is part of Olena. // @@ -24,37 +23,38 @@ // exception does not however invalidate any other reasons why the // executable file might be covered by the GNU General Public License. - -#include <libgen.h> #include <iostream> #include <mln/core/image/image2d.hh> -#include <mln/core/alias/neighb2d.hh> - #include <mln/io/pbm/load.hh> -#include <mln/io/ppm/save.hh> - -#include <mln/value/rgb8.hh> #include <mln/value/int_u16.hh> +#include <mln/draw/box_plain.hh> +#include <mln/pw/all.hh> +#include <mln/core/image/dmorph/image_if.hh> +#include <mln/data/convert.hh> +#include <mln/literal/colors.hh> -#include <mln/draw/box.hh> -#include <mln/draw/line.hh> +#include <scribo/text/recognition.hh> +#include <scribo/debug/usage.hh> + +#include <scribo/core/document.hh> +#include <scribo/core/component_set.hh> +#include <scribo/core/object_links.hh> +#include <scribo/core/object_groups.hh> #include <scribo/text/extract_lines.hh> -#include <scribo/debug/char_space_image.hh> -#include <scribo/debug/usage.hh> +#include <scribo/io/xml/save.hh> const char *args_desc[][2] = { - { "input.pbm", "A binary image. 'False' for objects, 'True'\ -for the background." }, {0, 0} }; + int main(int argc, char* argv[]) { using namespace scribo; @@ -62,23 +62,26 @@ int main(int argc, char* argv[]) if (argc != 3) return scribo::debug::usage(argv, - "Find text lines using left/right validation and display x-height in a binarized article.", - "input.pbm out.ppm", + "Show text lines", + "input.pbm output.ppm", args_desc); - trace::entering("main"); - image2d<bool> input; mln::io::pbm::load(input, argv[1]); typedef image2d<scribo::def::lbl_type> L; - - line_set<L> line = scribo::text::extract_lines(input, c8()); - - image2d<value::rgb8> output = scribo::debug::char_space_image(input, line); - mln::io::ppm::save(output, argv[2]); + line_set<L> lines = text::extract_lines(input, c8()); + + // Color image with line bboxes. + { + image2d<value::rgb8> output = data::convert(value::rgb8(), input); + for_all_lines(l, lines) + if (lines(l).is_textline()) + mln::draw::box(output, lines(l).bbox(), literal::red); + mln::io::ppm::save(output, argv[2]); + } trace::exiting("main"); } -- 1.5.6.5

13 years, 8 months

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

Olena-patches