DVS3D technology: Levels of Details

Levels of Detail (LOD) is a technique for rendering optimization based on distances between models and view positions in 3D environments. It is a useful technique for rendering large-volume graphics and processing a large amount of detailed data.

The features of naked vision determine that the visual effects are almost the same for high-definition and low-definition models when they are far away from the view position. However, high-definition models require a large number of triangles to represent their geometrical shapes while low-definition models only require a few.

Therefore, the renderer adjusts the resolution of models according to distances between models and view positions to reduce CPU-to-GPU throughput generated by triangle patches, thereby providing high-speed real-time rendering, as shown in Figure 

1. Figure 1 LOD theory demonstration

 

DVS3D adopts a smooth-transparentizing LOD modeling technique that allows transparent fading and transitions in overlapped areas of images with two different resolutions. This provides a better visual effect for users.

As shown in Figure 2, the parameter Visible indicates the distances between the model and camera that determine the disappearance and appearance of the model (switch between low-definition and high-definition models). The parameter Fade indicates the distances between the model and camera that determine the gradual transparency of the model.

 

Figure 2 Smooth-transparentizing LOD

 

The LOD technique adopted by DVS3D, which is a static LOD technique, needs to cache all low-definition and high-definition models in the memory.

For large-scale models, models far away from view positions are not displayed regardless of the resolutions so as to provide the best rendering effect.