SoLevelOfDetail -

#include <Inventor/nodes/SoLevelOfDetail.h>

Inherits SoGroup.

The SoLevelOfDetail class is used to choose a child based on projected size.

A level-of-detail mechanism is typically used by application programmers to assist the library in speeding up the rendering.

The way a level-of-detail mechanism works is basically like this:

Several versions of varying complexity of the same geometry / shape is provided by the application programmer in sorted order from 'most complex' to 'least complex' (where 'complex' in this context should be taken to mean more or less detailed in the number of polygons / shapes used for rendering it).

The run-time rendering system then, upon scenegraph traversal, will on-the-fly calculate either the distance from the camera to the 3D-model in question, or the number of pixels in the screen projection of the 3D-model. This value is then used to decide which version of the model to actually render: as the model is moved farther away from the camera, a less detailed version of the model is used. And vice versa, as the model moves closer to the camera, more and more detailed versions of it are rendered.

This is under many different circumstances a very effective way to let the application programmer assist to profoundly optimize the rendering of her 3D-scene.

There is of course a trade-off with the level-of-detail technique: more versions of the same 3D model means the scenegraph will use up more application memory resources. Also, generating the set of less and less detailed versions of a 3D model from the original is often not a trivial task to do properly. The process is often assisted by software like what Systems in Motion offers in their <a href='http://www.rational-reducer.com>Rational Reducer package.

The SoLevelOfDetail node implements the 'projected size' variety level-of-detail technique (as opposed to the 'object distance' technique, as done by the SoLOD node).

The node works by comparing the current projected size of the smallest rectangle covering the bounding box of it's child geometry.

Along with this set of models of the same shape, a specification of when to switch between them is also provided.

Example scenegraph layout:

  LevelOfDetail {
     screenArea [ 2000, 500, 50 ]

DEF version-0 Separator { # most complex / detailed / heavy version of subgraph } DEF version-1 Separator { # less complex version of subgraph } DEF version-2 Separator { # even less complex version of subgraph } DEF version-3 Separator { # simplest / 'lightest' version of subgraph } }

The way the above sub-scenegraph would work would be the following: if the rectangular area around the model's projected bounding box covers more than 2000 pixels (meaning it will be up pretty close to the camera), the most complex version of the model (version-0) would be traversed (and rendered, of course). If the projected area would be between 500 and 2000 pixels, the version-1 model would be used. Ditto if the projected area was between 50 and 500 pixels, the version-2 version of the model would be used. Finally, if the projected bounding box area would be less than 50 square pixels, the presumably least detailed version of the modeled would be used.

(A common 'trick' is to let the last of the SoLevelOfDetail node's children be just an empty subgraph, so no geometry will be rendered at all if the model is sufficiently far away. This will of course have a positive effect on the total rendering time for the complete scenegraph.)

Note that the SoLevelOfDetail::screenArea vector will be influenced by preceding SoComplexity nodes in the following way: if SoComplexity::value is set from 0.0 up to 0.5, lower detail levels than normal will be selected for traversal. If SoComplexity::value is above 0.5, higher level details than normal will be used. An SoComplexity::value equal to 1.0 will cause the first child of SoLevelOfDetail to always be used.

As mentioned above, there is one other level-of-detail node in the Coin library: SoLOD. The difference between that one and this is just that instead of projected bounding box area, SoLOD uses the distance between the camera and the object to find out when to switch between the different model versions.

Using SoLOD is faster, since figuring out the projected bounding box area needs a certain amount of calculations. But using SoLevelOfDetail is often 'better', in the sense that it's really a model's size and visibility in the viewport that determines whether we could switch to a less complex version without losing enough detail that it gives a noticable visual degradation.

FILE FORMAT/DEFAULTS:

    LevelOfDetail {
        screenArea 0
    }

See also: SoLOD

Public Member Functions

Static Public Member Functions

Public Attributes

Protected Member Functions

Static Protected Member Functions

Friends

SoLevelOfDetail::SoLevelOfDetail (void)

SoLevelOfDetail::SoLevelOfDetail (int numchildren)

SoLevelOfDetail::~SoLevelOfDetail () [protected, virtual]

SoType SoLevelOfDetail::getClassTypeId (void) [static]

SoType SoLevelOfDetail::getTypeId (void) const [virtual]

  void foo(SoNode * node)
  {
    if (node->getTypeId() == SoFile::getClassTypeId()) {
      SoFile * filenode = (SoFile *)node;  // safe downward cast, knows the type
    }
    else if (node->getTypeId().isOfType(SoGroup::getClassTypeId())) {
      SoGroup * group = (SoGroup *)node;  // safe downward cast, knows the type
    }
  }

const SoFieldData ** SoLevelOfDetail::getFieldDataPtr (void) [static, protected]

const SoFieldData * SoLevelOfDetail::getFieldData (void) const [protected, virtual]

void SoLevelOfDetail::initClass (void) [static]

void SoLevelOfDetail::doAction (SoAction * action) [virtual]

void SoLevelOfDetail::callback (SoCallbackAction * action) [virtual]

void SoLevelOfDetail::GLRender (SoGLRenderAction * action) [virtual]

void SoLevelOfDetail::rayPick (SoRayPickAction * action) [virtual]

void SoLevelOfDetail::getBoundingBox (SoGetBoundingBoxAction * action) [virtual]

void SoLevelOfDetail::audioRender (SoAudioRenderAction * action) [virtual]

void SoLevelOfDetail::notify (SoNotList * nl) [virtual]

SoMFFloat SoLevelOfDetail::screenArea

Generated automatically by Doxygen for Coin from the source code.