Skip to content

CorrectlyRenderTranslucentGeometry

Repository source: CorrectlyRenderTranslucentGeometry


Description

  • Contributed by: Lars Friedrich

Correctly rendering translucent geometry with OpenGL-functionality in the background (as in the case of VTK) requires non-intersecting polygons and depth-sorted traversal. In general these requirements are not satisfied as the inherent order of scene traversal is object-based. Using a method, namely depth peeling, presented by NVIDIA in 2001 Interactive Order-Independent Transparency, shadow mapping (multi-pass rendering) in conjunction with alpha test can be consulted to achieve correct blending of the rendered objects in the frame buffer.

VTK implements this feature since November 2006 as described in the VTK WIKI (Francois Bertel). Unfortunately depth peeling has several OpenGL extension, context and driver requirements (but also runs on Mesa) which restrict the approach's usage to modern GPUs. Usually this feature slows down the rendering process depending on the configuration (occlusion ratio and maximum number of iterative peels).

However if depth peeling is not available on a certain machine, depth sorting can be accomplished on the CPU using DepthSortPolyData. This is usually much slower than the GPU-implementation and furthermore brings additional restrictions with it (e.g. poly data must be merged within one set).

This example program generates a set of intersecting (overlapping) spheres that have transparency properties. The program automatically checks whether depth peeling is supported or not. If depth peeling is not supported, CPU-based depth sorting is used. In addition the program tries to determine an average frame rate for the scene.

The following image shows the spheres arrangement (and view position) and compares the different render modes: no special translucency treatment, CPU depth sorting and GPU depth peeling.

Program Usage

./CorrectlyRenderTranslucentGeometry Theta Phi MaximumPeels OcclusionRatio ForceDepthSortingFlag DoNotUseAnyDepthRelatedAlgorithmFlag

Theta ... spheres' THETA resolution

Phi ... spheres' PHI resolution

MaximumPeels ... maximum number of depth peels (multi-pass rendering) for depth peeling mode

OcclusionRatio ... occlusion ratio for depth peeling mode (0.0 for a perfect rendered image, >0.0 for a non-perfect image which is expected to be slower)

ForceDepthSortingFlag ... force depth sorting even if depth peeling is supported

DoNotUseAnyDepthRelatedAlgorithmFlag ... neither use depth peeling nor depth sorting - just render as usual

Example calls:

./CorrectlyRenderTranslucentGeometry 100 100 50 0.1 0 0 ... will render the spheres using depth peeling if available (depth sorting otherwise)

./CorrectlyRenderTranslucentGeometry 100 100 50 0.1 1 0 ... will render the spheres using depth sorting even if depth peeling is available

./CorrectlyRenderTranslucentGeometry 100 100 50 0.1 0 1 ... will render the spheres using neither depth peeling nor depth sorting

Resultant frame rates show that depth peeling is usually much faster than the CPU-implementation, however, it will slow down the rendering process due to internal multi-pass rendering.

Question

If you have a question about this example, please use the VTK Discourse Forum

Code

CorrectlyRenderTranslucentGeometry.cxx

#include <vtkActor.h>
#include <vtkAppendPolyData.h>
#include <vtkCamera.h>
#include <vtkDepthSortPolyData.h>
#include <vtkNamedColors.h>
#include <vtkNew.h>
#include <vtkPolyDataMapper.h>
#include <vtkProperty.h>
#include <vtkRenderWindow.h>
#include <vtkRenderWindowInteractor.h>
#include <vtkRenderer.h>
#include <vtkSmartPointer.h>
#include <vtkSphereSource.h>
#include <vtkTimerLog.h>
#include <vtkTransform.h>

namespace {

/**
 * Generate a bunch of overlapping spheres within one poly data set:
 * one big sphere evenly surrounded by four small spheres that intersect the
 * centered sphere.
 * @param theta sphere sampling resolution (THETA)
 * @param phi sphere sampling resolution (PHI)
 * @return the set of spheres within one logical poly data set
 **/
vtkSmartPointer<vtkAppendPolyData> GenerateOverlappingBunchOfSpheres(int theta,
                                                                     int phi);

/**
 * Setup the rendering environment for depth peeling (general depth peeling
 * support is requested).
 * @see IsDepthPeelingSupported()
 * @param renderWindow a valid openGL-supporting render window
 * @param renderer a valid renderer instance
 * @param maxNoOfPeels maximum number of depth peels (multi-pass rendering)
 * @param occulusionRation the occlusion ration (0.0 means a perfect image,
 * >0.0 means a non-perfect image which in general results in faster rendering)
 * @return TRUE if depth peeling could be set up
 */
bool SetupEnvironmentForDepthPeeling(vtkRenderWindow* renderWindow,
                                     vtkRenderer* renderer, int maxNoOfPeels,
                                     double occlusionRatio);

/**
 * Find out whether this box supports depth peeling. Depth peeling requires
 * a variety of openGL extensions and appropriate drivers.
 * @param renderWindow a valid openGL-supporting render window
 * @param renderer a valid renderer instance
 * @param doItOffscreen do the test off screen which means that nothing is
 * rendered to screen (this requires the box to support off screen rendering)
 * @return TRUE if depth peeling is supported, FALSE otherwise (which means
 * that another strategy must be used for correct rendering of translucent
 * geometry, e.g. CPU-based depth sorting)
 */
bool IsDepthPeelingSupported(vtkRenderWindow* renderWindow,
                             vtkRenderer* renderer, bool doItOffScreen);
} // namespace

/**
 * Example application demonstrating correct rendering of translucent geometry.
 * It will automatically detect whether depth peeling is supported by the
 * hardware and software, and will apply depth peeling if possible. Otherwise
 * a fallback strategy is used: depth sorting on the CPU.
 * <br>Usage:
 * [ProgramName] Theta Phi MaximumPeels OcclusionRatio ForceDepthSortingFlag
 * DoNotUseAnyDepthRelatedAlgorithmFlag
 * <br>
 * Theta ... spheres' THETA resolution <br>
 * Phi ... spheres' PHI resolution <br>
 * MaximumPeels ... maximum number of depth peels (multi-pass rendering) for
 * depth peeling mode <br>
 * OcclusionRatio ... occlusion ratio for depth peeling mode (0.0 for a
 * perfect rendered image, >0.0 for a non-perfect image which is expected to
 * be slower) <br>
 * ForceDepthSortingFlag ... force depth sorting even if depth peeling is
 * supported <br>
 * DoNotUseAnyDepthRelatedAlgorithmFlag ... neither use depth peeling nor
 * depth sorting - just render as usual
 */
int main(int argc, char* argv[])
{
  if (argc != 7)
  {
    std::cerr << "Usage: " << argv[0] << " Theta Phi MaximumPeels "
              << "OcclusionRatio ForceDepthSortingFlag "
              << "DoNotUseAnyDepthRelatedAlgorithmFlag" << std::endl;
    std::cerr << "100 100 50 0.1 0 0" << std::endl;
    return EXIT_FAILURE;
  }

  vtkNew<vtkNamedColors> colors;

  int theta = atoi(argv[1]);
  int phi = atoi(argv[2]);
  bool forceDepthSort = atoi(argv[5]) == 1;
  bool withoutAnyDepthThings = atoi(argv[6]) == 1;

  // Generate a translucent sphere poly data set that partially overlaps:
  auto translucentGeometry = GenerateOverlappingBunchOfSpheres(theta, phi);

  // generate a basic Mapper and Actor
  vtkNew<vtkPolyDataMapper> mapper;
  mapper->SetInputConnection(translucentGeometry->GetOutputPort());

  vtkNew<vtkActor> actor;
  actor->SetMapper(mapper);
  actor->GetProperty()->SetOpacity(0.5); // translucent !!!
  actor->GetProperty()->SetColor(colors->GetColor3d("Crimson").GetData());
  actor->RotateX(-72); // put the objects in a position where it is easy to see
                       // different overlapping regions

  // Create the RenderWindow, Renderer and RenderWindowInteractor
  vtkNew<vtkRenderer> renderer;
  vtkNew<vtkRenderWindow> renderWindow;
  renderWindow->SetSize(600, 400);
  renderWindow->AddRenderer(renderer);
  renderWindow->SetWindowName("CorrectlyRenderTranslucentGeometry");

  vtkNew<vtkRenderWindowInteractor> renderWindowInteractor;
  renderWindowInteractor->SetRenderWindow(renderWindow);

  // Add the actors to the renderer, set the background and size
  renderer->AddActor(actor);
  renderer->SetBackground(colors->GetColor3d("SlateGray").GetData());

  // Setup view geometry
  renderer->ResetCamera();
  renderer->GetActiveCamera()->Zoom(2.2); // so the object is larger
  renderWindow->Render();

  // Answer the key question: Does this box support GPU Depth Peeling?
  bool useDepthPeeling = IsDepthPeelingSupported(renderWindow, renderer, true);
  std::cout << "DEPTH PEELING SUPPORT: " << (useDepthPeeling ? "YES" : "NO")
            << std::endl;

  int success = EXIT_SUCCESS;

  // Use depth peeling if available and not explicitly prohibited, otherwise we
  // use manual depth sorting
  std::cout << std::endl << "CHOSEN MODE: ";
  if (useDepthPeeling && !forceDepthSort && !withoutAnyDepthThings) // GPU
  {
    int maxPeels = atoi(argv[3]);
    double occulusionRatio = atof(argv[4]);

    std::cout << "*** DEPTH PEELING ***" << std::endl;
    // Setup GPU depth peeling with configured parameters
    success = !SetupEnvironmentForDepthPeeling(renderWindow, renderer, maxPeels,
                                               occulusionRatio);
  }
  else if (!withoutAnyDepthThings) // CPU
  {
    std::cout << "*** DEPTH SORTING ***" << std::endl;
    // Setup CPU depth sorting filter
    vtkNew<vtkDepthSortPolyData> depthSort;
    depthSort->SetInputConnection(translucentGeometry->GetOutputPort());
    depthSort->SetDirectionToBackToFront();
    depthSort->SetVector(1, 1, 1);
    depthSort->SetCamera(renderer->GetActiveCamera());
    depthSort->SortScalarsOff(); // do not really need this here
    // Bring it to the mapper's input
    mapper->SetInputConnection(depthSort->GetOutputPort());
    depthSort->Update();
  }
  else
  {
    std::cout << "*** NEITHER DEPTH PEELING NOR DEPTH SORTING ***" << std::endl;
  }

  // Initialize interaction
  renderWindowInteractor->Initialize();

  // Check the average frame rate when rotating the actor
  int endCount = 100;
  vtkNew<vtkTimerLog> clock;
  // Set a user transform for successively rotating the camera position
  vtkNew<vtkTransform> transform;
  transform->Identity();
  transform->RotateY(2.0); // rotate 2 degrees around Y-axis at each iteration
  auto camera = renderer->GetActiveCamera();
  double camPos[3]; // camera position
  // Start test
  clock->StartTimer();
  for (int i = 0; i < endCount; i++)
  {
    camera->GetPosition(camPos);
    transform->TransformPoint(camPos, camPos);
    camera->SetPosition(camPos);
    renderWindow->Render();
  }
  clock->StopTimer();
  double frameRate = (double)endCount / clock->GetElapsedTime();
  std::cout << "AVERAGE FRAME RATE: " << frameRate << " fps" << std::endl;

  // Start interaction
  renderWindowInteractor->Start();

  return success;
}

namespace {

vtkSmartPointer<vtkAppendPolyData> GenerateOverlappingBunchOfSpheres(int theta,
                                                                     int phi)
{
  vtkNew<vtkAppendPolyData> appendData;

  for (int i = 0; i < 5; i++)
  {
    vtkNew<vtkSphereSource> sphereSource;
    sphereSource->SetThetaResolution(theta);
    sphereSource->SetPhiResolution(phi);
    sphereSource->SetRadius(0.5); // all spheres except the center
                                  // one should have radius = 0.5
    switch (i)
    {
    case 0:
      sphereSource->SetRadius(1);
      sphereSource->SetCenter(0, 0, 0);
      break;
    case 1:
      sphereSource->SetCenter(1, 0, 0);
      break;
    case 2:
      sphereSource->SetCenter(-1, 0, 0);
      break;
    case 3:
      sphereSource->SetCenter(0, 1, 0);
      break;
    case 4:
      sphereSource->SetCenter(0, -1, 0);
      break;
    }
    sphereSource->Update();
    appendData->AddInputConnection(sphereSource->GetOutputPort());
  }

  return appendData;
}

bool SetupEnvironmentForDepthPeeling(vtkRenderWindow* renderWindow,
                                     vtkRenderer* renderer, int maxNoOfPeels,
                                     double occlusionRatio)
{
  if (!renderWindow || !renderer)
    return false;

  // 1. Use a render window with alpha bits (as initial value is 0 (false)):
  renderWindow->SetAlphaBitPlanes(true);

  // 2. Force to not pick a framebuffer with a multisample buffer
  // (as initial value is 8):
  renderWindow->SetMultiSamples(0);

  // 3. Choose to use depth peeling (if supported) (initial value is 0 (false)):
  renderer->SetUseDepthPeeling(true);

  // 4. Set depth peeling parameters
  // - Set the maximum number of rendering passes (initial value is 4):
  renderer->SetMaximumNumberOfPeels(maxNoOfPeels);
  // - Set the occlusion ratio (initial value is 0.0, exact image):
  renderer->SetOcclusionRatio(occlusionRatio);

  return true;
}

bool IsDepthPeelingSupported(vtkRenderWindow* renderWindow,
                             vtkRenderer* renderer, bool doItOffScreen)
{

  bool success = true;

  // Save original renderer / render window state
  bool origOffScreenRendering = renderWindow->GetOffScreenRendering() == 1;
  bool origAlphaBitPlanes = renderWindow->GetAlphaBitPlanes() == 1;
  int origMultiSamples = renderWindow->GetMultiSamples();
  bool origUseDepthPeeling = renderer->GetUseDepthPeeling() == 1;
  int origMaxPeels = renderer->GetMaximumNumberOfPeels();
  double origOcclusionRatio = renderer->GetOcclusionRatio();

  // Activate off screen rendering on demand
  renderWindow->SetOffScreenRendering(doItOffScreen);

  // Setup environment for depth peeling (with some default parametrization)
  success = success &&
      SetupEnvironmentForDepthPeeling(renderWindow, renderer, 100, 0.1);

  // Do a test render
  renderWindow->Render();

  // Check whether depth peeling was used
  success = success && renderer->GetLastRenderingUsedDepthPeeling();

  // recover original state
  renderWindow->SetOffScreenRendering(origOffScreenRendering);
  renderWindow->SetAlphaBitPlanes(origAlphaBitPlanes);
  renderWindow->SetMultiSamples(origMultiSamples);
  renderer->SetUseDepthPeeling(origUseDepthPeeling);
  renderer->SetMaximumNumberOfPeels(origMaxPeels);
  renderer->SetOcclusionRatio(origOcclusionRatio);

  return success;
}

} // namespace

CMakeLists.txt

cmake_minimum_required(VERSION 3.12 FATAL_ERROR)

project(CorrectlyRenderTranslucentGeometry)

find_package(VTK COMPONENTS 
  CommonColor
  CommonCore
  CommonSystem
  CommonTransforms
  FiltersCore
  FiltersHybrid
  FiltersSources
  InteractionStyle
  RenderingContextOpenGL2
  RenderingCore
  RenderingFreeType
  RenderingGL2PSOpenGL2
  RenderingOpenGL2
)

if (NOT VTK_FOUND)
  message(FATAL_ERROR "CorrectlyRenderTranslucentGeometry: Unable to find the VTK build folder.")
endif()

# Prevent a "command line is too long" failure in Windows.
set(CMAKE_NINJA_FORCE_RESPONSE_FILE "ON" CACHE BOOL "Force Ninja to use response files.")
add_executable(CorrectlyRenderTranslucentGeometry MACOSX_BUNDLE CorrectlyRenderTranslucentGeometry.cxx )
  target_link_libraries(CorrectlyRenderTranslucentGeometry PRIVATE ${VTK_LIBRARIES}
)
# vtk_module_autoinit is needed
vtk_module_autoinit(
  TARGETS CorrectlyRenderTranslucentGeometry
  MODULES ${VTK_LIBRARIES}
)

Download and Build CorrectlyRenderTranslucentGeometry

Click here to download CorrectlyRenderTranslucentGeometry and its CMakeLists.txt file. Once the tarball CorrectlyRenderTranslucentGeometry.tar has been downloaded and extracted,

cd CorrectlyRenderTranslucentGeometry/build

If VTK is installed:

cmake ..

If VTK is not installed but compiled on your system, you will need to specify the path to your VTK build:

cmake -DVTK_DIR:PATH=/home/me/vtk_build ..

Build the project:

make

and run it:

./CorrectlyRenderTranslucentGeometry

WINDOWS USERS

Be sure to add the VTK bin directory to your path. This will resolve the VTK dll's at run time.