Information Technology Reference
In-Depth Information
(a)
(b)
Fig. 16
Frames interpolated using (a) the proposed method and (b) the conventional method
Fig. 16 illustrates sample interpolation results for synthetically generated
Rotat-
ing Cube
sequence using both the conventional method and the proposed method.
The figure reveals that conventional translational motion model is not capable of
handling complex motion types, and hence the interpolated frames are significantly
distorted. On the other hand, the rigid body motion model utilized by the FRUC-RM
algorithm yields qualitatively better frames.
The utilization of 3D motion parameters during frame interpolation enables ren-
dering of any desired number of frames in between the successive frames of the
original multi-view sequences, by the following motion parameter interpolation
scheme:
=
R
k
n
+ 1
,
a
,
R
i
k
n
+ 1
t
,
t
i
=
(24)
R
b
,
i
=
R
b
1
θ
b
,
a
b
,
k
n
+ 1
−
t
b
,
i
=
1
t
b
,
k
n
+ 1
−
k
= 1
,
2
,..,
n
.
where
n
denotes the number of frames to be interpolated and
k
represents the index
of an interpolation frame. Fig. 17 shows three interpolation frames obtained via the
above equations for
Rotating Cube
sequence.
