Adaptive Blocksize Observation Model for MRF modeling of Motion field

A suitable initialization motion field can reduce the computational cost needed by the stochastic process in MRF based motion estimation. The initialization is highly needed if deterministic optimization is used. The initialization process is commonly based on the intensity constant assumption, which is called an implicit constraint. An objective in our novel model is to add the implicit constraint into the observation model. This model yields a motion vector that has been tested as the only best candidate being proposed on a given point. The rest of the optimization process is assisted by the explicit constraint. Results indicate that the aperture problem can also be solved.