Image Processing Reference
In-Depth Information
x 10
4
a
b
1.42
x 10
4
1.32
1.22
1.4
1.3
1.12
x
-
MI
j
0
+
MI
j
x 10
4
1.2
1.419
1.1
1.319
y
1.219
x
1.119
0
-
MI
i
+
MI
i
y
Fig. 5.3 Example of inherent 3D holoscopic spatial correlation: (a) autocorrelation function for a
3D holoscopic image; and (b) projection onto x and y axis, showing the high correlation between
points spaced of about one micro-image size (MI)
Although there is a resemblance between 3D holoscopic video and 2D video
(as both are captured by an ordinary 2D sensor), a more careful analysis reveals
inherent correlations that are not exploited by state-of-the-art 2D video coding
solutions. Notably, in the spatial domain, a significant correlation between neigh-
boring micro-images can be identified through the autocorrelation function, as
illustrated in Fig.
5.3
. In particular, it can be seen that the pixel correlation in 3D
holoscopic content is not as smooth as in conventional 2D video content. A periodic
structure is evidenced by the autocorrelation function whose period is approxi-
mately one micro-image size (represented in each direction by MI
j
and MI
i
in
Fig.
5.3b
). Additionally, it should also be noted that each micro-image itself has
some degree of inter-pixel redundancy, as is also common in 2D images.
Some coding schemes in the literature have proposed to represent the 3D
holoscopic image by a stack of their composing micro-images, which can be
interpreted as a pseudo volumetric image (PVI) [
8
,
9
]—if stacking micro-images
are done along the third dimension—or a pseudo video sequence (PVS) [
10
,
11
]—if
stacking is done along the temporal dimension.
Briefly, consider a 3D holoscopic image, HI, as illustrated in Fig.
5.4a
, captured
using a rectangular-packed square-based micro-lens array with resolution
MLA
n
MLA
m
and micro-image resolution of MI
j
MI
i
. Each micro-image,
MI
k
, in the PVI or PVS representation can be obtained from HI at the position
k
(
x
,
y
) represents the
pixel positions inside MI
k
. Alternatively, the holoscopic image can be expressed in
terms of its micro-images by the array in (
5.2
).
¼
(
k
n
,
k
m
) in the micro-lens array, as given by (
5.1
), where x
¼
MI
k
¼
HI
k
n
MI
j
þ
x
,
k
m
MI
i
þ
y
ð
5
1
Þ
:
HI
¼
½
MI
k
MLA
n
MLA
m
ð
5
2
Þ
: