View as Slides

Converting Volumes to Meshes with VTK

Deborah Schmidt
Helmholtz Imaging | MDC Berlin
Sep 25, 2024
Slides available at https://ida-mdc.github.io/workshop-visualization/tutorial-mesh-conversion/

Introduction

  • VTK provides robust tools for generating meshes from volumes, including labelmaps and raw pixel data.
  • FlyingEdges3D is a highly efficient algorithm for generating iso-surfaces from volumetric data.
  • VTK supports advanced options for mesh optimization, including smoothing and decimation.

Solution to run conversion code automatically

To simplify and unify solutions used in our tutorials, we use Album, a tool for capturing and sharing specific software use cases in dedicated virtual environments.

Click the solution box next to this text and follow the displayed usage instructions to run the solution either from command line or graphical interface.

Conversion of 3D pixel datasets into meshes with VTK

visualization:pixel-volumes-to-meshes-vtk:0.2.0

This solution uses vtkFlyingEdges3D to generate meshes from pixel volumes, including optional smoothing and decimation of vertices, and can optionally display the result.
Tags: bioinformatics, neuroscience, data-analysis

Close solution

Conversion of 3D pixel datasets into meshes with VTK

visualization:pixel-volumes-to-meshes-vtk:0.2.0

This solution uses vtkFlyingEdges3D to generate meshes from pixel volumes, including optional smoothing and decimation of vertices, and can optionally display the result.
Tags: bioinformatics, neuroscience, data-analysis

  • License: MIT
  • Solution creator(s): Deborah Schmidt
Usage:
album install visualization:pixel-volumes-to-meshes-vtk:0.2.0
album run visualization:pixel-volumes-to-meshes-vtk:0.2.0
    --input # (file): Path to the 3D TIFF file containing region data or to a folder containing 3D masks. [Default: ] [Required: Yes]
    --output # (directory): Path to the place where the calculated meshes should be stored. [Default: ] [Required: Yes]
    --detect_labelmap # (boolean): Process int type images as labelmaps. [Default: 1] [Required: No]
    --contour_value # (float): In case continuous datasets are provided, this value can be set to generate meshes at a specific contour value. Otherwise it will be calculated between min and max of the dataset range. [Default: ] [Required: No]
    --smoothing_iterations # (integer): Smoothing iterations, usually between 10 and 50. More iterations typically lead to smoother results. Set to 0 to skip smoothing. [Default: 5] [Required: No]
    --smoothing_pass_band # (float): Smoothing pass band, determines the amount of smoothing. Lower values result in more smoothing. Usually between 0.0 and 2.0. [Default: 0.5] [Required: No]
    --smoothing_feature_angle # (float): The feature angle is the angle between the normals of adjacent polygons. A smaller feature angle will preserve edges. Value in degrees. [Default: 50.0] [Required: No]
    --decimate_percentage # (float): How much to decimate vertices after mesh generation (0-1, 0 means do not decimate). [Default: 0.5] [Required: No]
    --preserve_topology # (boolean): Preserve topology during mesh decimation. [Default: 0] [Required: No]
    --show_meshes # (boolean): Whether to display the generated mesh plots or not. [Default: 0] [Required: No]
    --colormap # (string): Name of the matplotlib colormap to use for coloring meshes. [Default: viridis] [Required: No]
    --exclude_zero # (boolean): Exclude 0 as a label when processing the volumes. [Default: 1] [Required: No]

Cite the following resources if you use this solution:

Dependencies

Link to full source code

314    dependencies={'environment_file': """channels:
315  - conda-forge
316dependencies:
317  - python=3.9
318  - pandas=1.5.3
319  - matplotlib-base=3.7.0
320  - numpy=1.24.2
321  - vtk=9.2.6
322  - numpy-stl=3.0.1
323  - opencv=4.8.0
324  - pip
325  - pip:
326    - vtkplotlib==2.1.0
327    - imagecodecs
328    - tifffile==2023.9.26
329"""}

Volume conversion using vtkFlyingEdges3D

  • Add zero border around dataset
  • Load volumetric image using vtkImageData
  • Generate meshes using vtkFlyingEdges3D algorithm
  • Set threshold between foreground and background via GenerateValues (int numContours, double rangeStart, double rangeEnd)

Link to full source code

 72        # pad the data to get closed meshes at the border of the image
 73        data = np.pad(data, ((1, 1), (1, 1), (1, 1)), 'constant', constant_values=(0,))
 74
 75        # numpy array to vtk array to vtkImageData
 76
 77        depth, height, width = data.shape
 78        imageData = vtk.vtkImageData()
 79        imageData.SetDimensions(width, height, depth)
 80        imageData.SetSpacing((1, 1, 1))
 81
 82        np_dtype = data.dtype
 83        vtk_dtype = vtk.VTK_UNSIGNED_CHAR
 84        if np_dtype == 'float32':
 85            vtk_dtype = vtk.VTK_FLOAT
 86        elif np_dtype == 'float64':
 87            vtk_dtype = vtk.VTK_DOUBLE
 88        elif np_dtype == 'uint8':
 89            vtk_dtype = vtk.VTK_UNSIGNED_CHAR
 90
 91        vtk_array = numpy_support.numpy_to_vtk(num_array=data.ravel(order='C'), deep=True, array_type=vtk_dtype)
 92        imageData.GetPointData().SetScalars(vtk_array)
 93
 94        # apply flying threshold algorithm to generate mesh from volume
 95        discrete = vtk.vtkFlyingEdges3D()
 96        discrete.SetInputData(imageData)
 97        discrete.SetValue(1, contour)
 98        discrete.Update()
 99
100        next_input = discrete.GetOutput()

Smoothing

  • Number of iterations: Controls how many smoothing passes are applied.
  • Feature Edge Smoothing: Ensures that important edges are preserved while smoothing.
  • Boundary Smoothing: Smooths out the mesh along the boundaries.

Link to full source code

103            # apply smoothing
104            smoother = vtk.vtkWindowedSincPolyDataFilter()
105            smoother.SetInputData(next_input)
106            smoother.SetNumberOfIterations(smoothing_iterations)
107            smoother.BoundarySmoothingOn()
108            smoother.FeatureEdgeSmoothingOn()
109            smoother.SetFeatureAngle(smoothing_feature_angle)
110            smoother.SetPassBand(smoothing_pass_band)
111            smoother.NonManifoldSmoothingOn()
112            smoother.NormalizeCoordinatesOn()
113            smoother.Update()
114            next_input = smoother.GetOutput()

Decimate geometry

  • Target Reduction: Defines the percentage reduction in the number of polygons (e.g., 0.5 for 50% reduction).
  • Preserve Topology: Ensures that the overall shape and connectivity of the mesh are maintained during decimation.

Link to full source code

117            # apply decimation of geometry
118            decimate = vtk.vtkDecimatePro()
119            decimate.SetInputData(next_input)
120            decimate.SetTargetReduction(decimate_percentage)
121            if preserve_topology:
122                decimate.PreserveTopologyOn()
123            decimate.Update()
124
125            next_input = decimate.GetOutput()

Write STL file

The final step in this process is to write the generated mesh to an STL file. STL files are widely used for 3D printing, simulations, and visualizations in tools like Blender.

Link to full source code

155        # write binary STL file
156        writer = vtk.vtkSTLWriter()
157        writer.SetFileName(path)
158        writer.SetInputData(next_input)
159        writer.SetFileTypeToBinary()
160        writer.Write()
2024 | HELMHOLTZ IMAGINGExit slides