lsm/PdfPig
1
0
Fork 0
mirror of https://github.com/UglyToad/PdfPig.git synced 2026-03-10 00:23:29 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

update RecursiveXYCut to use DlaOptions

Browse Source
This commit is contained in:
BobLd
2020-05-23 20:07:43 +01:00
committed by Eliot Jones
parent 5362a335f5
commit 1438fec741
1 changed files with 98 additions and 59 deletions
Download Patch File Download Diff File Expand all files Collapse all files

157
src/UglyToad.PdfPig.DocumentLayoutAnalysis/PageSegmenter/RecursiveXYCut.cs
View File

@@ -7,8 +7,9 @@
using System.Linq;
using UglyToad.PdfPig.Geometry;
/// <inheritdoc />
/// <summary>
/// The recursive X-Y cut is a top-down page segmentation technique that decomposes a document
/// The recursive X-Y cut is a top-down page segmentation technique that decomposes a document
/// recursively into a set of rectangular blocks. This implementation leverages bounding boxes.
/// https://en.wikipedia.org/wiki/Recursive_X-Y_cut
/// <para>See 'Recursive X-Y Cut using Bounding Boxes of Connected Components' by Jaekyu Ha, Robert M.Haralick and Ihsin T. Phillips</para>
@@ -20,81 +21,73 @@
/// </summary>
public static RecursiveXYCut Instance { get; } = new RecursiveXYCut();
/// <inheritdoc />
/// <summary>
/// Get the blocks.
/// <para>Uses 'minimumWidth' = 0, 'dominantFontWidthFunc' = Mode(Width), 'dominantFontHeightFunc' = 1.5 x Mode(Height)</para>
/// Get the blocks using default options values.
/// </summary>
/// <param name="pageWords">The words in the page.</param>
/// <returns></returns>
public IReadOnlyList<TextBlock> GetBlocks(IEnumerable<Word> pageWords)
/// <param name="words">The page's words to segment into <see cref="TextBlock"/>s.</param>
/// <returns>The <see cref="TextBlock"/>s generated by the Recursive X-Y cut method.</returns>
public IReadOnlyList<TextBlock> GetBlocks(IEnumerable<Word> words)
{
return GetBlocks(pageWords, 0);
return GetBlocks(words, new RecursiveXYCutOptions());
}
/// <inheritdoc />
/// <summary>
/// Get the blocks.
/// <para>Uses 'dominantFontWidthFunc' = Mode(Width), 'dominantFontHeightFunc' = 1.5 x Mode(Height)</para>
/// Get the blocks using options values.
/// </summary>
/// <param name="pageWords">The words in the page.</param>
/// <param name="minimumWidth">The minimum width for a block.</param>
public IReadOnlyList<TextBlock> GetBlocks(IEnumerable<Word> pageWords, double minimumWidth)
/// <param name="words">The page's words to segment into <see cref="TextBlock"/>s.</param>
/// <param name="options">The <see cref="RecursiveXYCutOptions"/> to use.</param>
/// <returns>The <see cref="TextBlock"/>s generated by the Recursive X-Y cut method.</returns>
public IReadOnlyList<TextBlock> GetBlocks(IEnumerable<Word> words, DlaOptions options)
{
return GetBlocks(pageWords, minimumWidth,
(letters) =>
if (options is RecursiveXYCutOptions ryxcOptions)
{
if (words?.Any() != true)
{
var widths = letters.Select(x => Math.Max(Math.Round(x.Width, 3), Math.Round(x.GlyphRectangle.Width, 3)));
var mode = widths.Mode();
if (double.IsNaN(mode) || mode == 0)
{
mode = widths.Average();
}
return mode;
},
(letters) =>
{
var heights = letters.Select(x => Math.Round(x.GlyphRectangle.Height, 3));
var mode = heights.Mode();
if (double.IsNaN(mode) || mode == 0)
{
mode = heights.Average();
}
return mode * 1.5;
return EmptyArray<TextBlock>.Instance;
}
);
return GetBlocks(words,
ryxcOptions.MinimumWidth,
ryxcOptions.DominantFontWidthFunc,
ryxcOptions.DominantFontHeightFunc,
ryxcOptions.WordSeparator,
ryxcOptions.LineSeparator);
}
else
{
throw new ArgumentException("Options provided must be of type " + nameof(RecursiveXYCutOptions) + ".", nameof(options));
}
}
/// <summary>
/// Get the blocks.
/// </summary>
/// <param name="pageWords">The words in the page.</param>
/// <param name="minimumWidth">The minimum width for a block.</param>
/// <param name="dominantFontWidth">The dominant font width.</param>
/// <param name="dominantFontHeight">The dominant font height.</param>
public IReadOnlyList<TextBlock> GetBlocks(IEnumerable<Word> pageWords, double minimumWidth,
double dominantFontWidth, double dominantFontHeight)
{
return GetBlocks(pageWords, minimumWidth, k => dominantFontWidth, k => dominantFontHeight);
}
/// <summary>
/// Get the blocks.
/// </summary>
/// <param name="pageWords">The words in the page.</param>
/// <param name="words">The words in the page.</param>
/// <param name="minimumWidth">The minimum width for a block.</param>
/// <param name="dominantFontWidthFunc">The function that determines the dominant font width.</param>
/// <param name="dominantFontHeightFunc">The function that determines the dominant font height.</param>
public IReadOnlyList<TextBlock> GetBlocks(IEnumerable<Word> pageWords, double minimumWidth,
/// <param name="wordSeparator"></param>
/// <param name="lineSeparator"></param>
private IReadOnlyList<TextBlock> GetBlocks(IEnumerable<Word> words, double minimumWidth,
Func<IEnumerable<Letter>, double> dominantFontWidthFunc,
Func<IEnumerable<Letter>, double> dominantFontHeightFunc)
Func<IEnumerable<Letter>, double> dominantFontHeightFunc,
string wordSeparator, string lineSeparator)
{
if (pageWords.Count() == 0) return EmptyArray<TextBlock>.Instance;
// Filter out white spaces
words = words.Where(w => !string.IsNullOrWhiteSpace(w.Text));
if (!words.Any())
{
return EmptyArray<TextBlock>.Instance;
}
XYLeaf root = new XYLeaf(pageWords); // Create a root node.
XYLeaf root = new XYLeaf(words); // Create a root node.
XYNode node = VerticalCut(root, minimumWidth, dominantFontWidthFunc, dominantFontHeightFunc);
if (node.IsLeaf)
{
return new List<TextBlock> { new TextBlock((node as XYLeaf).GetLines()) };
return new List<TextBlock> { new TextBlock((node as XYLeaf).GetLines(wordSeparator), lineSeparator) };
}
else
{
@@ -102,7 +95,7 @@
if (leaves.Count > 0)
{
return leaves.Select(l => new TextBlock(l.GetLines())).ToList();
return leaves.Select(l => new TextBlock(l.GetLines(wordSeparator), lineSeparator)).ToList();
}
}
@@ -114,7 +107,7 @@
Func<IEnumerable<Letter>, double> dominantFontHeightFunc, int level = 0)
{
// Order words left to right
var words = leaf.Words.Where(w => !string.IsNullOrWhiteSpace(w.Text)).OrderBy(w => w.BoundingBox.Normalise().Left).ToArray();
var words = leaf.Words.OrderBy(w => w.BoundingBox.Normalise().Left).ToArray();
if (words.Length == 0)
{
@@ -123,7 +116,7 @@
// Create new leaf with non-whitespace words.
leaf = new XYLeaf(words);
if (leaf.CountWords() <= 1 || leaf.BoundingBox.Width <= minimumWidth)
{
// We stop cutting if
@@ -195,7 +188,7 @@
}
if (i == wordsCount - 1) projectionProfile.Add(currentProjection);
}
var newLeavesEnums = projectionProfile.Select(p => leaf.Words.Where(w =>
{
// Get words that are contained in each projection profiles
@@ -203,7 +196,7 @@
return normalisedBB.Left >= p.LowerBound && normalisedBB.Right <= p.UpperBound;
}));
var newLeaves = newLeavesEnums.Where(e => e.Count() > 0).Select(e => new XYLeaf(e));
var newLeaves = newLeavesEnums.Where(e => e.Any()).Select(e => new XYLeaf(e));
var newNodes = newLeaves.Select(l => HorizontalCut(l, minimumWidth,
dominantFontWidthFunc, dominantFontHeightFunc, level)).ToList();
@@ -221,7 +214,7 @@
Func<IEnumerable<Letter>, double> dominantFontHeightFunc, int level = 0)
{
// Order words bottom to top
var words = leaf.Words.Where(w => !string.IsNullOrWhiteSpace(w.Text)).OrderBy(w => w.BoundingBox.Normalise().Bottom).ToArray();
var words = leaf.Words.OrderBy(w => w.BoundingBox.Normalise().Bottom).ToArray();
if (words.Length == 0)
{
@@ -303,7 +296,7 @@
return normalisedBB.Bottom >= p.LowerBound && normalisedBB.Top <= p.UpperBound;
}));
var newLeaves = newLeavesEnums.Where(e => e.Count() > 0).Select(e => new XYLeaf(e));
var newLeaves = newLeavesEnums.Where(e => e.Any()).Select(e => new XYLeaf(e));
var newNodes = newLeaves.Select(l => VerticalCut(l, minimumWidth,
dominantFontWidthFunc, dominantFontHeightFunc, level)).ToList();
@@ -335,5 +328,51 @@
return value >= LowerBound && value <= UpperBound;
}
}
/// <summary>
/// Recursive X-Y cut page segmenter options.
/// </summary>
public class RecursiveXYCutOptions : PageSegmenterOptions
{
/// <summary>
/// The minimum width for a block.
/// <para>Default value is 1.</para>
/// </summary>
public double MinimumWidth { get; set; } = 1;
/// <summary>
/// The function that determines the dominant font width.
/// <para>Default value is the mode of the block's letters width.
/// If the mode is not available, the average is used.</para>
/// </summary>
public Func<IEnumerable<Letter>, double> DominantFontWidthFunc { get; set; } =
(letters) =>
{
var widths = letters.Select(x => Math.Max(Math.Round(x.Width, 3), Math.Round(x.GlyphRectangle.Width, 3)));
var mode = widths.Mode();
if (double.IsNaN(mode) || mode == 0)
{
mode = widths.Average();
}
return mode;
};
/// <summary>
/// The function that determines the dominant font height.
/// <para>Default value is the mode of the block's letters height times 1.5.
/// If the mode is not available, the average is used.</para>
/// </summary>
public Func<IEnumerable<Letter>, double> DominantFontHeightFunc { get; set; } =
(letters) =>
{
var heights = letters.Select(x => Math.Round(x.GlyphRectangle.Height, 3));
var mode = heights.Mode();
if (double.IsNaN(mode) || mode == 0)
{
mode = heights.Average();
}
return mode * 1.5;
};
}
}
}
}