Graphics Reference
In-Depth Information
DLP Board
Processor
Projecting Lens
Memory
DMD
Optics
Shaping Lens
Time
Color Wheel
Condension Lens
Screen
Light Source
(a)
(b)
Figure 1.11
(a) DLP projecting technology. (b) Single-chip DLP projection mechanism
errors. These three strategies are presented in the following sections. Pan et al. (2004) have
extensively studied the use of color pattern(s) (RGB) and 3-CCD camera. According to their
technique, one single color pattern is used, and the data acquisition is fast. If binary or gray-
level patterns are used, they must be switched and projected rapidly so that they are captured
in a short period. Rusinkiewicz et al. proposed to switch patterns by software (Rusinkiewicz
and Levoy, 2001; Rusinkiewicz et al., 2002). To reach fast image switching, Zhang and Yau
(2007) proposed to take advantage of the projection mechanism of the single-chip digital-
light-processing (DLP) technology. According to their approach, three primary color channels
are projected sequentially and repeatedly. This allows capture of three color channel images
separately using a synchronized DLP projector device with a digital camera.
A color wheel is a circular disk that spins rapidly. It is composed of R, G, and B filters that
color the white light once it passes from in front. Color lights are thus generated. The digital
micro-mirror synchronized with the color light, reflects it, and produces three R, G, and B
color channel images. Human perception cannot differentiate individual channels as a result
of the projection speed. Instead color images are seen. Three phase-shifted sinusoidal patterns
are encoded as three primary color channels, R, G, and B of a color image. Three patterns are
sent to the single-chip DLP projector from which color filters are removed. A CCD camera
is synchronized with the projector and captures each of the three color channels separately
into a computer. Unwrapping and phase-to-depth processing steps are applied to the sequence
of captured images to recover the depth information. Despite this high speed acquisition, fast
motion may still distort the reconstructed phase and hence the reconstructed 3D geometry.
Weise et al. (2007) proposed to estimate the error in phase shifting, which produces ripples
on the 3D reconstructed surface, and to compensate it. Also, this estimation can provide the
motion of the reconstructed 3D surface. Three-step phase shifting has been introduced in
Section 1.3 where a sinusoidal pattern is shifted by
2
3
to produce three patterns, the minimum
required to recover depth information:
I
1
(
x
,
y
)
=
I
0
(
x
,
y
)
+
I
mod
(
x
,
y
) cos (
φ
(
x
,
y
)
−
θ
)
,
I
2
(
x
,
y
)
=
I
0
(
x
,
y
)
+
I
mod
(
x
,
y
)
cos
(
φ
(
x
,
y
))
,
and
(1.22)
I
3
(
x
,
y
)
=
I
0
(
x
,
y
)
+
I
mod
(
x
,
y
) cos (
φ
(
x
,
y
)
+
θ
)
.
