Extract text and text rectangle coordinates from a Pdf file using itextsharp

Question

I'm trying to get all words and their location coordinates from a PDF file. I've succeeded using the Acrobat API on .NET. Now, I'm trying to get the same result using a free API, such as iTextSharp (the .NET version). I can get the text (line by line) with PRTokeniser, but I have no idea of how to get the coordinates of the line, let alone of each word.

Answer 1

My account is too new reply to Mark Storer's answer.

I wasn't able to directly use the LocationTextExtracationStrategy (I think I must be doing something wrong). When I used the LocationTextExtracationStrategy I was able to get the text but I couldn't figure out how to get the coords for each string (or line of strings).

I ended up subclassing the LocationTextExtracationStrategy and exposing the data I wanted because it does have it internally.

I also wanted it in .net... so here is a sloppy C# version of what I put together.

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;

using iTextSharp.text.pdf.parser;

namespace PdfHelper
{
    /// <summary>
    /// Taken from http://www.java-frameworks.com/java/itext/com/itextpdf/text/pdf/parser/LocationTextExtractionStrategy.java.html
    /// </summary>
    class LocationTextExtractionStrategyEx : LocationTextExtractionStrategy
    {
        private List<TextChunk> m_locationResult = new List<TextChunk>();
        private List<TextInfo> m_TextLocationInfo = new List<TextInfo>();
        public List<TextChunk> LocationResult 
        {
            get { return m_locationResult; }
        }
        public List<TextInfo> TextLocationInfo
        {
            get { return m_TextLocationInfo; }
        }

        /// <summary>
        /// Creates a new LocationTextExtracationStrategyEx
        /// </summary>
        public LocationTextExtractionStrategyEx()
        {
        }

        /// <summary>
        /// Returns the result so far
        /// </summary>
        /// <returns>a String with the resulting text</returns>
        public override String GetResultantText()
        {
            m_locationResult.Sort();

            StringBuilder sb = new StringBuilder();
            TextChunk lastChunk = null;
            TextInfo lastTextInfo = null;
            foreach (TextChunk chunk in m_locationResult)
            {
                if (lastChunk == null)
                {
                    sb.Append(chunk.Text);
                    lastTextInfo = new TextInfo(chunk);
                    m_TextLocationInfo.Add(lastTextInfo);
                }
                else
                {
                    if (chunk.sameLine(lastChunk))
                    {
                        float dist = chunk.distanceFromEndOf(lastChunk);

                        if (dist < -chunk.CharSpaceWidth)
                        {
                            sb.Append(' ');
                            lastTextInfo.addSpace();
                        }
                        //append a space if the trailing char of the prev string wasn't a space && the 1st char of the current string isn't a space
                        else if (dist > chunk.CharSpaceWidth / 2.0f && chunk.Text[0] != ' ' && lastChunk.Text[lastChunk.Text.Length - 1] != ' ')
                        {
                            sb.Append(' ');
                            lastTextInfo.addSpace();
                        }
                        sb.Append(chunk.Text);
                        lastTextInfo.appendText(chunk);
                    }
                    else
                    {
                        sb.Append('\n');
                        sb.Append(chunk.Text);
                        lastTextInfo = new TextInfo(chunk);
                        m_TextLocationInfo.Add(lastTextInfo);
                    }
                }
                lastChunk = chunk;
            }
            return sb.ToString();
        }

        /// <summary>
        /// 
        /// </summary>
        /// <param name="renderInfo"></param>
        public override void RenderText(TextRenderInfo renderInfo)
        {
            LineSegment segment = renderInfo.GetBaseline();
            TextChunk location = new TextChunk(renderInfo.GetText(), segment.GetStartPoint(), segment.GetEndPoint(), renderInfo.GetSingleSpaceWidth(), renderInfo.GetAscentLine(), renderInfo.GetDescentLine());
            m_locationResult.Add(location);
        }

        public class TextChunk : IComparable, ICloneable
        {
            string m_text;
            Vector m_startLocation;
            Vector m_endLocation;
            Vector m_orientationVector;
            int m_orientationMagnitude;
            int m_distPerpendicular;
            float m_distParallelStart;
            float m_distParallelEnd;
            float m_charSpaceWidth;

            public LineSegment AscentLine;
            public LineSegment DecentLine;

            public object Clone()
            {
                TextChunk copy = new TextChunk(m_text, m_startLocation, m_endLocation, m_charSpaceWidth, AscentLine, DecentLine);
                return copy;
            }

            public string Text
            {
                get { return m_text; }
                set { m_text = value; }
            }
            public float CharSpaceWidth
            {
                get { return m_charSpaceWidth; }
                set { m_charSpaceWidth = value; }
            }
            public Vector StartLocation
            {
                get { return m_startLocation; }
                set { m_startLocation = value; }
            }
            public Vector EndLocation
            {
                get { return m_endLocation; }
                set { m_endLocation = value; }
            }

            /// <summary>
            /// Represents a chunk of text, it's orientation, and location relative to the orientation vector
            /// </summary>
            /// <param name="txt"></param>
            /// <param name="startLoc"></param>
            /// <param name="endLoc"></param>
            /// <param name="charSpaceWidth"></param>
            public TextChunk(string txt, Vector startLoc, Vector endLoc, float charSpaceWidth, LineSegment ascentLine, LineSegment decentLine)
            {
                m_text = txt;
                m_startLocation = startLoc;
                m_endLocation = endLoc;
                m_charSpaceWidth = charSpaceWidth;
                AscentLine = ascentLine;
                DecentLine = decentLine;

                m_orientationVector = m_endLocation.Subtract(m_startLocation).Normalize();
                m_orientationMagnitude = (int)(Math.Atan2(m_orientationVector[Vector.I2], m_orientationVector[Vector.I1]) * 1000);

                // see http://mathworld.wolfram.com/Point-LineDistance2-Dimensional.html
                // the two vectors we are crossing are in the same plane, so the result will be purely
                // in the z-axis (out of plane) direction, so we just take the I3 component of the result
                Vector origin = new Vector(0, 0, 1);
                m_distPerpendicular = (int)(m_startLocation.Subtract(origin)).Cross(m_orientationVector)[Vector.I3];

                m_distParallelStart = m_orientationVector.Dot(m_startLocation);
                m_distParallelEnd = m_orientationVector.Dot(m_endLocation);
            }

            /// <summary>
            /// true if this location is on the the same line as the other text chunk
            /// </summary>
            /// <param name="textChunkToCompare">the location to compare to</param>
            /// <returns>true if this location is on the the same line as the other</returns>
            public bool sameLine(TextChunk textChunkToCompare)
            {
                if (m_orientationMagnitude != textChunkToCompare.m_orientationMagnitude) return false;
                if (m_distPerpendicular != textChunkToCompare.m_distPerpendicular) return false;
                return true;
            }

            /// <summary>
            /// Computes the distance between the end of 'other' and the beginning of this chunk
            /// in the direction of this chunk's orientation vector.  Note that it's a bad idea
            /// to call this for chunks that aren't on the same line and orientation, but we don't
            /// explicitly check for that condition for performance reasons.
            /// </summary>
            /// <param name="other"></param>
            /// <returns>the number of spaces between the end of 'other' and the beginning of this chunk</returns>
            public float distanceFromEndOf(TextChunk other)
            {
                float distance = m_distParallelStart - other.m_distParallelEnd;
                return distance;
            }

            /// <summary>
            /// Compares based on orientation, perpendicular distance, then parallel distance
            /// </summary>
            /// <param name="obj"></param>
            /// <returns></returns>
            public int CompareTo(object obj)
            {
                if (obj == null) throw new ArgumentException("Object is now a TextChunk");

                TextChunk rhs = obj as TextChunk;
                if (rhs != null)
                {
                    if (this == rhs) return 0;

                    int rslt;
                    rslt = m_orientationMagnitude - rhs.m_orientationMagnitude;
                    if (rslt != 0) return rslt;

                    rslt = m_distPerpendicular - rhs.m_distPerpendicular;
                    if (rslt != 0) return rslt;

                    // note: it's never safe to check floating point numbers for equality, and if two chunks
                    // are truly right on top of each other, which one comes first or second just doesn't matter
                    // so we arbitrarily choose this way.
                    rslt = m_distParallelStart < rhs.m_distParallelStart ? -1 : 1;

                    return rslt;
                }
                else
                {
                    throw new ArgumentException("Object is now a TextChunk");
                }
            }
        }

        public class TextInfo
        {
            public Vector TopLeft;
            public Vector BottomRight;
            private string m_Text;

            public string Text
            {
                get { return m_Text; }
            }

            /// <summary>
            /// Create a TextInfo.
            /// </summary>
            /// <param name="initialTextChunk"></param>
            public TextInfo(TextChunk initialTextChunk)
            {
                TopLeft = initialTextChunk.AscentLine.GetStartPoint();
                BottomRight = initialTextChunk.DecentLine.GetEndPoint();
                m_Text = initialTextChunk.Text;
            }

            /// <summary>
            /// Add more text to this TextInfo.
            /// </summary>
            /// <param name="additionalTextChunk"></param>
            public void appendText(TextChunk additionalTextChunk)
            {
                BottomRight = additionalTextChunk.DecentLine.GetEndPoint();
                m_Text += additionalTextChunk.Text;
            }

            /// <summary>
            /// Add a space to the TextInfo.  This will leave the endpoint out of sync with the text.
            /// The assumtion is that you will add more text after the space which will correct the endpoint.
            /// </summary>
            public void addSpace()
            {
                m_Text += ' ';
            }


        }
    }
}

I added a TextLocationInfo property which hands back a List of lines of text + the coords for those lines (upper left and lower right) which can be used to give you a bounding box.

I also saw something odd with my initial playing around. It looked like I got the same coords if I pulled the startPoint & endPoint from the baseline (I think the right thing to do, and the thing I did, was to pull those points from ascentLine and DecentLine). My initial pass I just used the baseline. Odd that I didn't see a difference in the resulting coords. So word to the wary... I'm not sure if the coords I'm providing are right... I just think they are/should be.

Answer 2

You'll want to use the com.itextpdf.text.pdf.parser package classes. They track the current transformation, color, font, and so forth.

Sadly, these classes weren't covered in the new book, so you're left with the JavaDoc, and mentally converting it all from Java to C#, which isn't much of a stretch.

So you'll want to plug a LocationTextExtractionStrategy into a PdfTextExtractor.

This will give you the strings and locations as they are presented in the pdf. It will be up to you to interpret that as words (and paragraphs if need be, ouch).

Keep in mind that PDF doesn't know anything about text layout. Every character can be placed individually. If someone were so inclined (and they'd have to be a few tacos short of a combo platter to do so) they could draw all the 'a's on a given page, then all the 'b's, and so forth.

More realistically, someone might draw all the text on the page that uses FontA, then everything for FontB, and so on. This can produce more efficient content streams. Keep in mind that italic and bold (and bold italic) are all separate fonts. If someone marks only part of a word as bold (or whatever), then that Logical Word is required to be broken up into at least two drawing commands.

But lots of folks just write out their text into PDF in logical order... which is Very Handy for folks who are stuck trying to parse it, but you Must Not Expect It. Because you will invariably run into some oddball that doesn't.