Use of Stereoscopic Photography to Distinguish Object Depth in Outdoor Scene
Chapter 3 The Geometry of Multiple Views
3.1 Depth Perception Methodology
Figure 2. A demonstration of binocular parallax. (a) Left image frame. (b) Right image frame.
In Figure 2(a) we can see that the signs appearing in the rear of the scene is
tended to left just like human’s left eye which has observed it or otherwise vice versa
as shown in Figure 2(b). The phenomenon of binocular parallax is the principal
theory of stereoscopic photography. We put the camera in two different places and
took the picture in turn. In other words, we can take a pair of pictures to simulate
binocular parallax effect. Binocular parallax effect depends on the distance between
two shots which we call it as stereo base line. We can get that the longer the stereo
base line the better the stereo effect for the view of a far way mountain. Figure 3 is a
demonstration of stereoscopic photograph taken by a camera.
Figure 3. A demonstration of stereoscopic photograph.
Some new and advantageous photographic technologies are being introduced
such as image base virtual reality, panama, and stereo photo. Many of them are
being implemented in the commercial digital camera, with the only exception of
stereo photo efficacy, which is being implemented only in some types of Pentax
cameras. We proposed a simplified method of taking stereo photo by introducing
the use of only one commercial digital camera. We propose two stages of works in
our experiment.
Use of Stereoscopic Photography to Distinguish Object Depth in Outdoor Scene
First stage, we can take serial photos with regular displacement along the slide
bar and make preprocesses if necessary. The steps of taking stereo photo are showed
as follows.
1. Estimate the distance of the main objects which you want to take the photo.
2. Determine the horizontal distance of the camera between two shots.
3. Carry out post processing to combine left and right image and take a look with
accessories.
Second stage, according to the different purposes, we can divide the works into
two parts. First part, for entertainment we can select photo-pairs and use
stereoscopic glasses to present with 3D scene. Second part, for application we
propose the model of lineal regression after training and validation. We can use the
results in real world for depth estimation.
The flowchart of stereoscopic photograph is shown in Figure 4.
The X-Z plane coordination of our proposed stereoscopic photograph camera
system is shown in Figure 5. Many important parameters are also shown in this
Figure.
x
lz
= fx (2.1)
x
rz b x f ( − )
= (2.2)
From Eq. 2.1 and Eq. 2.2, we obtain Eq. 2.3, as follows.
Figure 4. The flowchart of stereoscopic photograph.
x - x
l rz b x f
fx − ( − )
=
z
r
l
x
x
b x f fx
−
−
= − ( )
z d
= fb (2-3)
where d = x
l- x
r, is the disparity of two corresponding point.
Use of Stereoscopic Photography to Distinguish Object Depth in Outdoor Scene
Figure 5. Coordinate system in X-Z plane.
⎥ ⎥
⎥ ⎥
⎥ ⎥
⎦
⎤
⎢ ⎢
⎢ ⎢
⎢ ⎢
⎣
⎡
=
⎥ ⎥
⎥ ⎥
⎥ ⎥
⎦
⎤
⎢ ⎢
⎢ ⎢
⎢ ⎢
⎣
⎡
−
=
⎥ ⎥
⎥
⎦
⎤
⎢ ⎢
⎢
⎣
⎡
d fb
f z y
f z x
x x
fb f
z y
f z x
z y x
l l
r l
l l
(2.4)
Depth (z) goes to infinite as disparity (d) approaches to zero. As shown in Eq.
2.5 we can see that depth(z) is increased while disparity(d) is down if base line(b)
keeps a constant. We can use this property to estimate the focus of the camera by
Least Square Method.
z
0
lim
d