In the times of digital media, the processing of video sequences and movies
becomes more important every year. A particular challenge in the context of
video processing is the fact that objects typically move between subsequent
frames. Hence, algorithms developed for image processing cannot simply be
extended by the time dimension without significant loss of quality.
Evidently, motion information has to be taken into consideration, whenever
any computation takes place. A typical example for video processing is
deinterlacing:
|
|
Interlaced Image |
Deinterlaced Result |
-
Deinterlacing
In the era of HDTV the conversion of videos from deinterlaced to progressive
format becomes more and more important. The goal is thereby to recover the
even and odd lines that have intentionally been omitted in the transmission
process to achieve higher frame rates. To obtain optimal results, both spatial
and temporal information has to be exploited for this task. This requires
the use of accurate spatial interpolation methods as well as precise motion
estimation techniques that
carefully steer the temporal interpolation process.
In [1] we developed such an algorithm based on motion-compensated
anisotropic diffusion. It combines the quality of sophisticated directional
motion-adaptive interpolation schemes with the accuracy of recent motion estimators.
-
Framerate Conversion and Retiming
Another important application of motion information in the context of video
processing is the linear or nonlinear temporal resampling of recorded video
sequences (framerate conversion, retiming). To this end, additional frames
are computed by interpolation along the estimated motion trajectories.
Currently, a variant of our real-time algorithm presented in [2]
is used by the Rhythm and Hues Studios
in Hollywood for the retiming of action scenes. A recent movie that makes use
of our optic flow algorithms is
Fast and Furios.
-
M. Ghodstinat, A. Bruhn, J. Weickert:
Deinterlacing with motion-compensated anisotropic diffusion.
In D. Cremers, B. Rosenhahn, A. Yuille, F. Schmidt (Eds.):
Statistical and Geometrical Approaches to Visual Motion Analysis.
Lecture Notes in Computer Science, Vol. 5604, pp. 91-016, Springer,
Berlin, 2009.
-
A. Bruhn, J. Weickert:
Towards ultimate motion estimation: Combining highest accuracy with
real-time performance.
In Proc. Tenth IEEE International Conference on Computer Vision, Vol. 1,
749-755, IEEE Computer Society Press, 2005.
|