Extract Graphics

As of version 3.0, it is possible to extract the existing graphics on a page as a collection of ContentShape objects. This collection and its items can be navigated, inspected, modified and persisted. This allows you to actually change the PDF content.

Page.CreateShapes

The central method to extracting graphics is Page.CreateShapes. The following code copies all content except the images from an existing document to a new document:

static void Main(string[] args)
{
   using (FileStream fileIn = new FileStream(
      "in.pdf", FileMode.Open, FileAccess.Read))
   {
      Document pdfIn = new Document(fileIn);
      Document pdfOut = new Document();

      foreach ( Page page in pdfIn.Pages)
         pdfOut.Pages.Add(copy(page));

      using (FileStream fileOut = new FileStream(
         "out.pdf", FileMode.Create, FileAccess.Write))
      {
         pdfOut.Write(fileOut);
      }
   }
}

static Page copy(Page page)
{
   Page newPage = new Page(page.Width, page.Height);

   ShapeCollection shapes = page.CreateShapes();
   purge(shapes);
   newPage.Overlay.Add(shapes);

   return newPage;
}

static void purge(ShapeCollection shapes)
{
   for(int i=0; i<shapes.Count; i++)
   {
      Shape shape = shapes[i];
      if (shape is ImageShape) { shapes.RemoveAt(i); i--; }
      if (shape is ShapeCollection) purge(shape as ShapeCollection);
   }
}

Code sample: Copy content except images from an existing document to a new document

Shape types

Only the following shape types are returned:  

• Structure shapes: ShapeCollection, ClipShape and LayerShape
• Path shapes: FreeHandShape and LineShape
• Graphics shapes: TextShape and ImageShape

Transformations

The Transform property of each ContentShape is set to an object of type TransformCollection. The items inside this collection are always of type MatrixTransform. Each transformation inside a PDF document is preserved as a single MatrixTransform so they are not collapsed.

Restrictions

The following features are not preserved while extracting graphics since they are not yet supported by the Shape object model:  

• Complex transparencies will be reduced to a single opacity value of type double (ContentShape.Opacity).
• When modifying the TextShape.Text property, you need to be aware that if it uses a subsetted font, it may not be able to display the new text. This is the case if characters are used for which no representation exists in the subsetted font.
• Shading types FunctionBasedShading (Type1), FreeFormGouraudShaded (Type4), LatticeFormGouraudShaded (type5), CoonsPatchMeshes (type6) and TensorProductPatchMeshes (type7) shadings are not supported. These are mapped to no fill.
• The functions of shading types Axial Shading (type2) and RadialShading(type3) are converted to an array of color stops to achieve a similar visual effect.
• Not all PDF color spaces are supported. Especially CIE based color spaces, DeviceN, Indexed and Pattern. These are converted to RGB to achieve a similar visual effect.