Information Technology Reference
In-Depth Information
1
Introduction
Next generation hold-type panel displays emerged into the consumer market and the
demand for these items is expected to continue growing. In the near future, due to
their compact and aesthetic design, as well as high brightness, the hold type displays
are expected to completely replace Cathode Ray Tube (CRT) displays, with scan-
ning type properties. However, one of the main drawbacks of hold-type displays,
such as Liquid Crystal Displays (LCD) and Electro-Luminescence Displays (ELD),
is the motion blur which is caused by two major reasons [1]:
Long response time of crystal cell
: In moving images, pixel luminance changes
between two consecutive frames. Due to its long response time, the luminance
switch can not be finished within one frame period, which leads to the smearing
of object edges.
Constant pixel luminance over the entire frame period (hold-type)
:Humaneyeis
apt to move and track the moving objects while the successive frames are displayed.
In hold-type displays, when eyes move, many pixels are displayed on the same reti-
nal position during one frame period, which causes blurred image observed. The
faster the motion is, the more pixels are accumulated at the same retinal position
during one frame period and the more motion-blur is observed.
Therefore, for a better perception quality, the frame rate of the original video sig-
nal is required to be increased. Frame rate up-conversion (FRUC), or scan/field rate
up-conversion, which emerged as a result of this necessity, is a technique of increas-
ing the frame rate of a video signal by inserting interpolated frame(s) in-between
the successive frames of the original signal. The two most commonly used FRUC
techniques are frame repetition and motion compensation based methods. Frame
repetition generally yield inferior results, especially on complex (high depth varia-
tion and clutter) or dynamic (moving objects) scenes. Motion compensation based
approaches yield relatively more elegant results on these scenes with the cost of a
higher computational complexity and provided that accompanying motion estima-
tion and interpolation schemes provide satisfactorily accurate estimates. Although
all the methods in this category perform motion estimation, motion model parameter
interpolation and motion compensation steps, they exhibit a great variety in the way
the three steps are performed. Unlike coding approaches, which simply try to mini-
mize the residual error between adjacent frames, in FRUC algorithms it is crucial to
estimate the
correct
motion parameters.
2
Related Work
Most of the existing algorithms in the literature utilize block-based motion estima-
tion, [2] - [18], based on the assumption that all the pixels of an individual block
have the same
translational
motion. One of the key elements in the motion es-
timation for FRUC is the temporal search direction. In addition to the backward
directional motion estimation, which is mostly preferred by video coding tech-
niques, FRUC approaches also utilize forward directional motion estimation, [2],
