Enhanced Motion Blur Calculation with Optical Flow
Supervision:
Background:
Optical Flow is a method to compute an approximate motion field in an image
sequence. Normally the motion can only be detected at image
boundaries such as edges but in most applications one wants to have a dense
motion field, i.e. a motion vector at every point in the
image. Therefore a variational formulation of the optical flow problem is used
including the assumptions of a smooth motion field
and a constant gray value of a single point. The latter means that changes in
the illumination of the scene are ignored. This simple model
makes it possible to compute the optical flow very efficiently.
Computing motion blur is a complex task even for modern raytracers.
An alternative approach to temporal sampling is to used image
based techniques. These have e.g. been applied to stop motion animations.
The topic of this thesis will be to use information gained from
the optical flow algorithm, as described above,
to enhance image based motion blur computations.
Tasks:
-
Calculation of motion vectors with optical flow
-
Development and implementation of filtering methods for image based motion blur computations
-
Comparison to accurately motion blurred pictures using an external 3D program
Further Information:
Status:
Finished