[vtkusers] vtkAppendPolydata - wrong extents on multiple inputs

Nigel Mcfarlane Nigel.Mcfarlane at beds.ac.uk
Fri Aug 7 07:48:13 EDT 2015

Dear Dan,
I’m sorry, but I was too hasty in declaring that my problem was fixed – removing my AddScalars filter did not help after all, and neither did replacing it with vtkPolyDataNormals.  I have now stripped down the code to its simplest configuration.  The resulting scene has isosurfaces which are mispositioned and broken in a manner consistent with the images being read with the wrong dimensions.  Furthermore, if the readers are not manually updated, the program crashes.

There is also an error message:
ERROR: In C:\VTK\VTK-6.2.0\Common\ExecutionModel\vtkStreamingDemandDrivenPipeline.cxx, line 857
vtkCompositeDataPipeline (0132A818): The update extent specified in the information for output port 0 on algorithm vtkMetaImageReader(01321E58) is 0 67 0 72 0 9, which is outside the whole extent 0 69 0 58 0 10.

The results are ok if only one image is fed to appendPD.  If I update the dmcubes[i] manually, the results are also ok.

It looks as if the base of the pipeline only remembers one set of extents, which it then demands incorrectly from all the inputs.  Have I missed anything important in the code below?

  // Set filenames
  std::string dirname = CHICVIS"/ForthLungCropped20150703";
  std::string fname[3];
  fname[0] = dirname + "/2505ct/Tumor1Big/lesion2505-1mm.mhd";
  fname[1] = dirname + "/3108ct/Tumor1Big/lesion3108-1mm.mhd";
  fname[2] = dirname + "/170957ct/Tumor1Big/lesion1709-1mm.mhd";

  // Create array of image readers
  vtkMetaImageReader **reader;
  reader = new (vtkMetaImageReader *[3]);
  for (int i = 0; i < 3; i++)
    reader[i] = vtkMetaImageReader::New();

  for (int i = 0; i < 3; i++){

  // -----------------------
  // Create the VTK pipeline
  //  image0     image1      image2
  // dmcubes0   dmcubes1    dmcubes2
  //         \     |       /
  //        vtkAppendPolyData
  //          mapper/actor
  //           renderer
  // -----------------------
  vtkDiscreteMarchingCubes **dmcubes;
  dmcubes = new (vtkDiscreteMarchingCubes *[3]);
  for (int i = 0; i < 3; i++){
    dmcubes[i] = vtkDiscreteMarchingCubes::New();
    dmcubes[i]->SetValue(i, 255);

  vtkAppendPolyData *appendPD = vtkAppendPolyData::New();
  for (int i = 0; i < 3; i++)

  vtkPolyDataMapper *mapper = vtkPolyDataMapper::New();
  mapper = vtkPolyDataMapper::New();

  vtkActor *actor = vtkActor::New();


  // -------------------------------
  // Reset the camera and initialize
  // -------------------------------

Thank you
Nigel McFarlane
University of Bedfordshire UK

From: Dan Lipsa [mailto:dan.lipsa at kitware.com]
Sent: 06 August 2015 22:21
To: Nigel Mcfarlane <Nigel.Mcfarlane at beds.ac.uk>; vtkusers at vtk.org
Subject: Re: [vtkusers] vtkAppendPolydata - wrong extents on multiple inputs

> This request is propagated upstream and stops if a filter has its inputs older than its output.
I meant downstream. It has to check all filters as their properties might change which will affect the final result.

On Thu, Aug 6, 2015 at 5:12 PM Dan Lipsa <dan.lipsa at kitware.com<mailto:dan.lipsa at kitware.com>> wrote:
The pipeline executive compares the modification time (vtkObject::GetMTime()) for the output of a filter against each of its inputs. If any of the inputs is newer, the filter is updated. It may be that your filter does not update the time of its output properly. This request is propagated upstream and stops if a filter has its inputs older than its output. I would setup a breakpoint and look at the first update for your filters.

On Thu, Aug 6, 2015 at 9:20 AM Nigel Mcfarlane <Nigel.Mcfarlane at beds.ac.uk<mailto:Nigel.Mcfarlane at beds.ac.uk>> wrote:
Dear vtkusers,

I am creating a visual pipe in vtk 6.2.0 and I am having problems connecting vtkAppendPolyData to multiple inputs which start from images of different dimensions.  The visual pipe converts a set of input images into isosurfaces, adds scalars for colour and transparency, then merges them into one polydata mesh so they can be depth sorted before rendering.  Its structure looks like this:

//     image1       image2       image3
//       |            |            |
//    DMCubes1     DMCubes2     DMCubes3
//       |            |            |
//  AddScalars1  AddScalars2  AddScalars3
//             \      |      /
//              \     |     /
//               \    |    /
//            vtkAppendPolyData
//                    |
//           vtkDepthSortPolyData
//                    |
//               Mapper/Actor
//                    |
//                 Renderer

The images are produced by vtkMetaImageReader, the “DMCubes” filters are vtkDiscreteMarchingCubes, and the”AddScalars” filters are my own code.

The problem is that this produces an image in which the dimensions/extents of two of the images have clearly been set incorrectly.  I don’t know much about the way the vtk pipeline works, but I suppose that an information request for the extents of the images travels up and back down the pipeline, so the extents get set to that of the last input, and then applied to all the inputs.  The problem disappears if the input branches are first independently updated by calling a manual Update() on each of the AddScalars filters, but it returns if one of the DMCubes filters is modified, which triggers another general update cycle.

Is there any way to set up this pipeline so that it does not require a manual update when the image inputs are modified?

Nigel McFarlane
University of Bedfordshire
Powered by www.kitware.com<http://www.kitware.com>

Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html

Please keep messages on-topic and check the VTK FAQ at: http://www.vtk.org/Wiki/VTK_FAQ

Search the list archives at: http://markmail.org/search/?q=vtkusers

Follow this link to subscribe/unsubscribe:
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/vtkusers/attachments/20150807/4c852738/attachment-0001.html>

More information about the vtkusers mailing list