Image Processing Reference
In-Depth Information
Fig. 4.10 Blending in view extraction
Fig. 4.11 Block matching
algorithm to estimate
disparity between MIs
reducing block artifacts by smoothing the transitions between adjacent patches. It
also blurs the parts of the picture that are out of the depth plane that is in focus
(closer or further), which eliminates the mismatch artifacts (described in the
previous section) and provides the same effect as in a classic 2D photograph with
limited depth of field.
4.2.4.5 Disparity Map
A disparity estimation method is proposed in [
8
] in order to obtain the depth of the
objects inside each MI. It is based on a block matching algorithm (illustrated in
Fig.
4.11
) with the following steps. A square patch
P
is first selected in the center of
the current MI with coordinates (
Px
,
Py
), and a second patch
P
0
is selected in a
neighbor MI (e.g., right and/or bottom) with coordinates
P
0
x
k
, with
S
the size of the MI and
k
the disparity between the two MIs. The similitude
between
P
and
P
0
is computed (e.g., using normalized cross correlation) for values
of
k
from 1 up to the maximum disparity value possible. The value of
k
providing
the maximum similarities between
P
and
P
0
corresponds to the disparity value of the
current MI. This value in number of pixels corresponds to the adequate patch size to
be used for the view extraction. The resulting extracted viewpoint image is full-
focused, as each patch size is adapted to the depth of the objects in the MI.
¼
Px
þ
S
þ
