Graphics Reference
In-Depth Information
Fig. 6.16
Eye-gaze corrected images using
a
N
-camera versus
b
confident camera recoloring
setup offline, but a built-in camera setup into the screen would avoid this procedure
due to fixed calibration parameters. Our software framework runs on an Intel Xeon
2.8GHz, equipped with 2GB system memory and an NVIDIA GeForce 8800GTX
graphics card. Communication with the GPU is done through OpenGL, and it is
programmed with the high-level GPU language Cg.
6.6.1 Visual Quality
Final quality results using N-camera view recoloring are shown in Fig.
6.16
a, and the
results using confident camera recoloring are depicted in Fig.
6.16
b. These results
are generated under moderate variable illumination conditions, but with a fixed set of
fine tuned practical system parameters summarized in Table
6.1
.InFig.
6.16
a, some
small artifacts along the ears and chin, together with minor ghosting around the neck,
can still be noticed due to limitations of the depth refinement. The results generated
with the confident camera recoloring are much more detailed and sharp, however
some minor artifacts can be noticed due to abrupt camera transitions in the color
map
H
v
. Nevertheless, the images maintain their integrity and are regarded as high
subjective visual quality, while they convincingly seem to be making eye contact
with the reader.
