Image Processing Reference
In-Depth Information
4.2.4 View Extraction
The connection between integral imaging and Super Multi-View can be drawn by
the mean of view extraction. Several methods can be used to extract a viewpoint
from a holoscopic image. A detailed overview of these methods is given in [
19
].
4.2.4.1 Basic Extraction
The most basic method to extract a viewpoint image from an integral image is to
take only one pixel from each MI, as illustrated in Fig.
4.7
. The angle of view can be
selected by varying the relative position of this pixel within the MI. The number of
angles of view available is equal to the resolution of the MI, and the resolution of
the extracted viewpoint images is equal to the total number of MIs. Hence there is a
tradeoff between the number and the resolution of the extracted views. This is the
main limitation of this method which generally provides views with a small
resolution and with pixelation artifacts.
4.2.4.2 View Selective Blending
The goal of this method is to extract intermediary points of view from neighbor
views extracted with the previous method. First these views are stacked onto each
other,andthentheyareblendedbyaveraging all the superimposed pixels as
illustrated in Fig.
4.8
. The resulting extracted image is the viewpoint centered
between the stacked views. A drift can be applied to each view in the stack
Fig. 4.7 Basic view
extraction method
