Image Processing Reference
In-Depth Information
MI
j
Micro-Image
Patch
PS
PP
i
MI
i
BASIC RENDERING
WEIGHTED BLENDING
PS
PP
j
Weight
Function
3D Holoscopic Image
MLA
n
PS
×
MLA
n
PS
×
MLA
m
PS
×
MLA
m
Generated
2D view
Generated
2D view
PS
×
MLA
n
MLA
m
a
b
c
Fig. 5.9 Generating a 2D view image [
15
] (Copyright
2013, IEEE): (a) 3D holoscopic input
image; (b) Basic Rendering algorithm [
7
]; and (c) Weighted Blending algorithm [
7
]
©
This algorithm makes use of the knowledge that there is redundant spatial
information in multiple micro-images, since each micro-image captures
overlapping areas of the scene. Therefore, the weighted blending consists in
averaging together all these overlapping regions across different micro-images,
weighting differently the overlapping pixels. As can be seen in Fig.
5.9c
, each
micro-image is overlapped with a shift of PS pixels (corresponding to the patch
size) to its neighboring micro-images in both horizontal and vertical directions.
Then the pixels in the same spatial position across various micro-images are
averaged by using a specific weighting function. This weighting function corre-
sponds to a bivariate Gaussian function (non-normalized) of size MI
j
MI
i
and
whose mean vector,
μ
, is determined by the patch center, and the covariance matrix,
Σ
, is defined depending on the patch size. Hence, the weight applied to a pixel in the
position p
(
p
i
,
p
j
) inside its corresponding micro-image is given by (
5.8
).
Similarly to the basic algorithm, by varying the position of the patch, it is possible
to extract several different 2D view images, each having a (PS
¼
MLA
n
)
(PS
MLA
m
) resolution.
Gaussian p
μ
¼
α
1
2
p
T
Σ
1
p
Weight p
ðÞ
¼
,
Σ
exp
ð
μ
Þ
ð
μ
Þ
ð
5
8
Þ
:
5.3.4.3 Common Parameters
The Basic Rendering and Weighted Blending algorithms can be controlled through
the following two main parameters:
1.
Size of the patch
(PS): Since a micro-image is captured by the corresponding
micro-lens in perspective projection geometry, placing two objects of the same
(real) size closer or farther from the micro-lens array will result in those objects