Triangulation-Based Approaches to Three-Dimensional Scene Reconstruction - 3D Computer Vision: Efficient Methods and Applications - page 61

Fig. 1.19 8 which are used to compute the disparity for the

two contour points marked by black circles according to the contour segment-based scheme given
by ( 1.110 ). The solid contour has been extracted from the right image ,the dashed contour from
the left image of the stereo pair

Two contour segments of length L S =

Fig. 1.20 3 adaptive thresh-

olds derived from the bounding box of each segmented blob, respectively. The contours are shown
on the d s

( a ) Disparity image obtained with binary threshold θ 0 = 7and N T =

0 plane of the three-dimensional plot. The disparities have been computed accord-
ing to the contour point-based scheme given by ( 1.107 ). ( b ) Disparity image of the same scene,
computed according to the contour segment-based scheme given by ( 1.110 ) with contour segment
length L S

=

0 . 5. It
is obvious that the contour segment-based method yields a significantly less noisy disparity image
than the contour point-based method

= 8, cross-correlation threshold θ corr = 0 . 7, and absolute slope threshold θ slope =

c (w (i)

− L S / 2 + i)

s (i)
1

=

1

for i = 1 ,...,L S (1.108)

1

c (w ( j)

− L S / 2 + j)

s (j)
2

=

2

for j = 1 ,...,L S . (1.109)

2

For an illustration of this contour segment extraction procedure, see Fig. 1.19 .The
contour segment-based disparity d c is then defined by

L S

¯ .

u (w ( j) u (w (i)

1
L S

− L S / 2 + i) − L S / 2 + i)

d c =

−¯

2 1

(1.110)

2 1

i = 1