The Ultrasound Visualization Pipeline - Scientific Visualization: Uncertainty, Multifield, Biomedical, and Scalable Visualization - page 280

We distinguish between local and global illumination models. Local illumination
models use gradients of the volumetric function instead of surface normals to evaluate
the diffuse and specular terms of the Phong illumination model [ 42 ]. While local
illumination models already reveal structures, global illumination methods result in a
more realistic appearance, which further supports spatial perception. While gradient-
based local illumination methods are faster to evaluate, gradient computation is
sensitive to noise and high frequencies, which are natural properties of ultrasound
data.
Recent works show that global illumination models based on gradient-free meth-
ods are suitable for rendering ultrasound volumes [ 62 , 76 ]. Ropinski et al. described
a volumetric lighting model which simulates scattering and shadowing [ 62 ]. They
use slice-based volume rendering from the view of the light source to calculate a
light volume and raycasting to render the final image (see Fig. 24.7 b). A perceptual
evaluation of the generated images indicates that the proposed model yields stronger
depth cues than gradient-based shading. Šoltészová et al. presented a single-pass
method for the simulation of light scattering in volumes [ 76 ]. Light transport is ap-
proximated using a tilted cone-shaped function which leaves elliptic footprints in the
opacity buffer during slice-based volume rendering. They use a slice-based renderer
with an additional opacity buffer. This buffer is incrementally blurred with an ellip-
tical kernel, and the algorithm generates a high-quality soft-shadowing effect (see
Fig. 24.7 c). The light position and direction can be interactivelymodified.While these
two techniques have been explicitly applied to 3D US data, the application of other
volumetric illumination models potentially also improves the visual interpretation
of 3D US data. Figure 24.8 shows a comparison of six different shading techniques
as applied to a 3D US scan of a human heart. While the first row of Fig. 24.8 shows
examples for the already addressed shading techniques, the second row shows three
alternative approaches. Figure 24.8 d incorporates scattering of light in volume data,
as proposed by Kniss et al. [ 37 ]. Their slicing technique allows textured slices to be
rendered from both light and viewing direction simultaneously. By sampling the in-
cident light frommultiple directions while updating the light's attenuation map, they
Fig. 24.7 a Diastole of the aortic valve on a modern ultrasound workstation using color-coding
basedondepth. b Rendering of 3D ultrasound of a human heart with shadowing from the work of
Ropinski et al. [ 62 ]and c rendered using the technique presented in the work of Šoltészová et al. [ 76 ]
© IEEEReprinted, with permission, from IEEE Pacific Visualization Symposium (PacificVis), 2010