lsm/PdfPig
1
0
Fork 0
mirror of https://github.com/UglyToad/PdfPig.git synced 2025-10-14 10:55:04 +08:00
Code Issues Actions 5 Packages Projects Releases Wiki Activity

allow oriented bounding box for TextLine

Browse Source
This commit is contained in:
BobLd
2020-05-23 19:43:27 +01:00
committed by Eliot Jones
parent bb94348127
commit 5b0b0a6db3
1 changed files with 278 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

295
src/UglyToad.PdfPig.DocumentLayoutAnalysis/TextLine.cs
View File

@@ -57,34 +57,295 @@
Words = words;
Text = string.Join(" ", words.Where(s => !string.IsNullOrWhiteSpace(s.Text)).Select(x => x.Text));
var normalisedBoundingBoxes = words.Select(x => NormaliseRectangle(x.BoundingBox)).ToList();
var minX = normalisedBoundingBoxes.Min(x => x.Left);
var minY = normalisedBoundingBoxes.Min(x => x.Bottom);
var maxX = normalisedBoundingBoxes.Max(x => x.Right);
var maxY = normalisedBoundingBoxes.Max(x => x.Top);
BoundingBox = new PdfRectangle(minX, minY, maxX, maxY);
if (words.All(x => x.TextOrientation == words[0].TextOrientation))
if (Words.Count == 1)
{
BoundingBox = Words[0].BoundingBox;
Text = Words[0].Text;
TextOrientation = words[0].TextOrientation;
}
else
{
TextOrientation = TextOrientation.Other;
var tempTextOrientation = words[0].TextOrientation;
if (tempTextOrientation != TextOrientation.Other)
{
foreach (var letter in words)
{
if (letter.TextOrientation != tempTextOrientation)
{
tempTextOrientation = TextOrientation.Other;
break;
}
}
}
switch (tempTextOrientation)
{
case TextOrientation.Horizontal:
BoundingBox = GetBoundingBoxH(words);
break;
case TextOrientation.Rotate180:
BoundingBox = GetBoundingBox180(words);
break;
case TextOrientation.Rotate90:
BoundingBox = GetBoundingBox90(words);
break;
case TextOrientation.Rotate270:
BoundingBox = GetBoundingBox270(words);
break;
case TextOrientation.Other:
default:
BoundingBox = GetBoundingBoxOther(words);
break;
}
Text = string.Join(Separator, words.Where(s => !string.IsNullOrWhiteSpace(s.Text)).Select(x => x.Text));
TextOrientation = tempTextOrientation;
}
}
private PdfRectangle NormaliseRectangle(PdfRectangle rectangle)
#region Bounding box
private PdfRectangle GetBoundingBoxH(IReadOnlyList<Word> words)
{
return new PdfRectangle(Math.Min(Math.Min(Math.Min(rectangle.TopLeft.X, rectangle.TopRight.X), rectangle.BottomLeft.X), rectangle.BottomRight.X),
Math.Min(Math.Min(Math.Min(rectangle.TopLeft.Y, rectangle.TopRight.Y), rectangle.BottomLeft.Y), rectangle.BottomRight.Y),
Math.Max(Math.Max(Math.Max(rectangle.TopLeft.X, rectangle.TopRight.X), rectangle.BottomLeft.X), rectangle.BottomRight.X),
Math.Max(Math.Max(Math.Max(rectangle.TopLeft.Y, rectangle.TopRight.Y), rectangle.BottomLeft.Y), rectangle.BottomRight.Y));
var blX = double.MaxValue;
var trX = double.MinValue;
var blY = double.MaxValue;
var trY = double.MinValue;
for (var i = 0; i < words.Count; i++)
{
var word = words[i];
if (word.BoundingBox.BottomLeft.X < blX)
{
blX = word.BoundingBox.BottomLeft.X;
}
if (word.BoundingBox.BottomLeft.Y < blY)
{
blY = word.BoundingBox.BottomLeft.Y;
}
var right = word.BoundingBox.BottomLeft.X + word.BoundingBox.Width;
if (right > trX)
{
trX = right;
}
if (word.BoundingBox.TopLeft.Y > trY)
{
trY = word.BoundingBox.TopLeft.Y;
}
}
return new PdfRectangle(blX, blY, trX, trY);
}
private PdfRectangle GetBoundingBox180(IReadOnlyList<Word> words)
{
var blX = double.MinValue;
var blY = double.MinValue;
var trX = double.MaxValue;
var trY = double.MaxValue;
for (var i = 0; i < words.Count; i++)
{
var word = words[i];
if (word.BoundingBox.BottomLeft.X > blX)
{
blX = word.BoundingBox.BottomLeft.X;
}
if (word.BoundingBox.BottomLeft.Y > blY)
{
blY = word.BoundingBox.BottomLeft.Y;
}
var right = word.BoundingBox.BottomLeft.X - word.BoundingBox.Width;
if (right < trX)
{
trX = right;
}
if (word.BoundingBox.TopRight.Y < trY)
{
trY = word.BoundingBox.TopRight.Y;
}
}
return new PdfRectangle(blX, blY, trX, trY);
}
private PdfRectangle GetBoundingBox90(IReadOnlyList<Word> words)
{
var b = double.MaxValue;
var r = double.MaxValue;
var t = double.MinValue;
var l = double.MinValue;
for (var i = 0; i < words.Count; i++)
{
var word = words[i];
if (word.BoundingBox.BottomLeft.X < b)
{
b = word.BoundingBox.BottomLeft.X;
}
if (word.BoundingBox.BottomRight.Y < r)
{
r = word.BoundingBox.BottomRight.Y;
}
var right = word.BoundingBox.BottomLeft.X + word.BoundingBox.Height;
if (right > t)
{
t = right;
}
if (word.BoundingBox.BottomLeft.Y > l)
{
l = word.BoundingBox.BottomLeft.Y;
}
}
return new PdfRectangle(new PdfPoint(b, l), new PdfPoint(t, l),
new PdfPoint(t, r), new PdfPoint(b, r));
}
private PdfRectangle GetBoundingBox270(IReadOnlyList<Word> words)
{
var t = double.MaxValue;
var b = double.MinValue;
var l = double.MaxValue;
var r = double.MinValue;
for (var i = 0; i < words.Count; i++)
{
var word = words[i];
if (word.BoundingBox.BottomLeft.X > b)
{
b = word.BoundingBox.BottomLeft.X;
}
if (word.BoundingBox.BottomLeft.Y < l)
{
l = word.BoundingBox.BottomLeft.Y;
}
var right = word.BoundingBox.BottomLeft.X - word.BoundingBox.Height;
if (right < t)
{
t = right;
}
if (word.BoundingBox.BottomRight.Y > r)
{
r = word.BoundingBox.BottomRight.Y;
}
}
return new PdfRectangle(new PdfPoint(b, l), new PdfPoint(t, l),
new PdfPoint(t, r), new PdfPoint(b, r));
}
private static PdfRectangle GetBoundingBoxOther(IReadOnlyList<Word> words)
{
var baseLinePoints = words.SelectMany(r => new[]
{
r.BoundingBox.BottomLeft,
r.BoundingBox.BottomRight,
}).ToList();
// Fitting a line through the base lines points
// to find the orientation (slope)
double x0 = baseLinePoints.Average(p => p.X);
double y0 = baseLinePoints.Average(p => p.Y);
double sumProduct = 0;
double sumDiffSquaredX = 0;
for (int i = 0; i < baseLinePoints.Count; i++)
{
var point = baseLinePoints[i];
var x_diff = point.X - x0;
var y_diff = point.Y - y0;
sumProduct += x_diff * y_diff;
sumDiffSquaredX += x_diff * x_diff;
}
double cos = 0;
double sin = 1;
if (sumDiffSquaredX > 1e-3)
{
// not a vertical line
double angleRad = Math.Atan(sumProduct / sumDiffSquaredX); // -π/2 ≤ θ ≤ π/2
cos = Math.Cos(angleRad);
sin = Math.Sin(angleRad);
}
// Rotate the points to build the axis-aligned bounding box (AABB)
var inverseRotation = new TransformationMatrix(
cos, -sin, 0,
sin, cos, 0,
0, 0, 1);
var transformedPoints = words.SelectMany(r => new[]
{
r.BoundingBox.BottomLeft,
r.BoundingBox.BottomRight,
r.BoundingBox.TopLeft,
r.BoundingBox.TopRight
}).Distinct().Select(p => inverseRotation.Transform(p));
var aabb = new PdfRectangle(transformedPoints.Min(p => p.X),
transformedPoints.Min(p => p.Y),
transformedPoints.Max(p => p.X),
transformedPoints.Max(p => p.Y));
// Rotate back the AABB to obtain to oriented bounding box (OBB)
var rotateBack = new TransformationMatrix(
cos, sin, 0,
-sin, cos, 0,
0, 0, 1);
// Candidates bounding boxes
var obb = rotateBack.Transform(aabb);
var obb1 = new PdfRectangle(obb.BottomRight, obb.BottomLeft, obb.TopLeft, obb.TopRight);
var obb2 = new PdfRectangle(obb.TopRight, obb.BottomRight, obb.BottomLeft, obb.TopLeft);
var obb3 = new PdfRectangle(obb.TopLeft, obb.TopRight, obb.BottomRight, obb.BottomLeft);
// Find the orientation of the OBB, using the baseline angle
// Assumes word order is correct
var firstWord = words[0];
var lastWord = words[words.Count - 1];
var baseLineAngle = Distances.Angle(firstWord.BoundingBox.BottomLeft, lastWord.BoundingBox.BottomRight);
double deltaAngle = Math.Abs(Distances.BoundAngle180(obb.Rotation - baseLineAngle));
double deltaAngle1 = Math.Abs(Distances.BoundAngle180(obb1.Rotation - baseLineAngle));
if (deltaAngle1 < deltaAngle)
{
deltaAngle = deltaAngle1;
obb = obb1;
}
double deltaAngle2 = Math.Abs(Distances.BoundAngle180(obb2.Rotation - baseLineAngle));
if (deltaAngle2 < deltaAngle)
{
deltaAngle = deltaAngle2;
obb = obb2;
}
double deltaAngle3 = Math.Abs(Distances.BoundAngle180(obb3.Rotation - baseLineAngle));
if (deltaAngle3 < deltaAngle)
{
obb = obb3;
}
return obb;
}
#endregion
/// <inheritdoc />
public override string ToString()
{