Image Processing Reference
In-Depth Information
extracted from a row u i inside the micro-image (see Fig. 5.4b ), it is possible to
define a row vector, r, in the holoscopic image HI, which is given by ( 5.5 ).
r m
ð
¼
k m , i
¼
u i
Þ ¼
k m
MI i þ
u i
ð
5
5
Þ
:
Thus, for a fixed row vector r( m , i ), the corresponding ray-space image, RI r , can be
given by ( 5.6 ), where x
( x , y ) corresponds to the relative pixel position inside this
ray-space image. Consequently, the resolution of RI r is, then, MLA n
¼
MI j .
RI r x
ðÞ ¼
HI r, y
MI j þ
x
ð
5
6
Þ
:
Similarly to the VI-based holoscopic image (HI VI ), an RI-based holoscopic image,
HI RI , can also be defined, which has the same resolution of the original 3D
holoscopic image HI if (and only if) the micro-images are square. Considering
square micro-images (i.e., MI j ¼
MI i ), this is possible by transposing each extracted
RI r and arranging it at the relative position given by r. In other words,
MLA m MI i
HI RI
RI r T
¼
ð
5
7
Þ
:
Through this format, the angular information in one direction (horizontal in this
case) across different micro-images is represented. Figure 5.7a shows an example of
the referred RI-based holoscopic image. It is possible to see in the enlargement
(Fig. 5.7b ) that there is also a significant amount of redundancy between neighboring
ray-space images. However, as can be seen in Fig. 5.8 with the autocorrelation
function, this redundancy is not equally distributed in horizontal and vertical
directions. Additionally, the periodic structure is also observed, as seen in
Fig. 5.8b , and is given, approximately, by the resolution in each direction
(represented by MLA n
MI j in Fig. 5.8b ).
Fig. 5.7 Ray-space image based representation: (a) RI-based holoscopic image; (b) enlargement
of 272
224 pixels showing each ray-space image in detail
Search WWH ::




Custom Search