sensor. Each elemental image contains a particular perspective
of the object. To reconstruct a true 3-D image the elemental
image array is displayed in a liquid crystal display (LCD) in
front of a microlens array or the so-called projection type II that
projects the elemental image array over a micro convex-mirror
array [5]. The reconstruction stage converges, or diverges, the
rays from each elemental image. The conventional pickup procedure
is shown in Fig. 1(a) and the display procedure for projection
type II is shown in Fig. 1(b).
Many studies have been performed in order to solve the
resolution limitation of II [6]–[10], increase depth of focus
[11]–[13], and many other analysis and challenges in II
[14]–[17], because of the interest of improving 3-D image
quality. Recently, the scaling problem in 3-D II was studied and
solved by increasing (or decreasing) the spatial ray sampling
rate of elemental image arrays in the pickup procedure [18].
Hereon, we call this technique the optical magnification (OM)
method. In the first stage of OM method, an adequate number
of sets of elemental-image arrays are picked up with a moving
array lenslet technique (MALT) [9]. The elemental-image
arrays are cut into separate elemental images and rearranged
into a single elemental-image array. The new elemental-image
array contains more elemental images than the original array.
Therefore, the spatial ray-sampling rate of the original elemental
image array is increased. The 3-D-scaled image is
reconstructed by displaying the new elemental image array
through a stationary lenslet array.
The need for picking up a high number of elemental-image arrays
makes the OM technique very complex. The number of arrays
increases as the square of the magnification factor. A major
problem of OM is the alignment of elemental images due to the
cut and rearrangement procedure. In this paper, we proposed
to reduce the complexity and alignment problem using digital
magnification (DM) processing.
Digital image magnification has been performed in 2-D
images since the beginning of computer graphics and image
processing [19], where simple interpolation algorithms were
applied to display larger images and re-sample digital images
using nearest neighbor or linear interpolation [20]. Nowadays
2-D digital magnification is more popular than optical magnifi-
cation because of its simplicity.
We follow the basic principle of 2-D DM that estimates unknown
pixel values from known values. Then we can calculate
the new elemental images required to over-sample the array
using original elemental images. This means that we can increase
the spatial ray sampling rate of the elemental image array
without lens movement. However, blurred 3-D integral images