Information Technology Reference
In-Depth Information
greatest hurdle to overcome is not the production of the computer generated 3D
holoscopic images themselves but the computational overhead required to attain
real-time speeds on high-resolution displays. 3D holoscopic images are by defini-
tion the re-integration of multiple disseminated intensity values to produce the at-
tribute of all-round viewing. To view a complete replayed volumetric scene from
any arbitrary viewing position and hence mimic the original scene exactly with
depth and parallax requires a high sampling rate that is dependent upon the scene
depth. Small pixel sizes are required to hold a satisfactory depth without compro-
mising viewer comfort and to generate a large enough display inevitably equates
to a very high number of calculated intensity values for each frame. Adding to the
complexities is the increased computations required when using spherical or
hexagonally packed microlens arrays that generate 3D holoscopic images with
omni-directional parallax. A less computationally severe option is to use semi-
cylindrical lens arrays that generate unidirectional 3D holoscopic images with
parallax in horizontal direction.
There has been a small amount of work that focused on the efficiency of the
execution time required for the generation of photo realistic 3D Holoscopic im-
ages. One of the techniques reported in literature is based on parallel group ren-
dering [33] where rather than rendering each perspective micro-image; each group
of parallel rays is rendered using orthographic projection. A slightly modified ver-
sion termed viewpoint vector rendering was later proposed to make the rendering
performance independent of the number of the micro-images [34]. In this method
each micro-image is assembled from a segmented area of the directional scenes.
Both techniques are based on rasterization rendering technique and hence do not
produce photo-realistic images.
A technique used to generate fast photo-realistic 3D Holoscopic images was re-
ported by Youssef
et. al
. [35]. The technique of accelerating ray tracing is to
reduce the number of intersection tests for shadow rays using a shadow cache al-
gorithm. The image-space coherence is analysed describing the relation between
rays and projected shadows in the scene rendered. Shadow cache algorithm has
been adapted in order to minimise shadow intersection tests in ray tracing of 3D
Holoscopic images. Shadow intersection tests make the majority of the intersec-
tion tests in ray tracing. The structure of the lenses and the camera model in the
3D Holoscopic image ray-tracing affects the way primary rays are spawned and
traced as well as the spatial coherence among successive rays. As a result various
pixel-tracing styles can be developed uniquely for 3D Holoscopic image ray trac-
ing to improve the image-space coherence and the performance of the shadow
cache algorithm. Examples of grouping of pixels tracing styles are shown in
figure 6. Acceleration of the photo-realistic 3D Holoscopic images generation us-
ing the image-space coherence information between shadows and rays in 3D holo-
scopic ray tracing has been achieved with up to 41% of time saving [35]. Also, it
has been proven that applying the new styles of pixel-tracing does not affect the
scalability of 3D Holoscopic image ray tracing running over parallel computers.
