Picture rate conversion using regression based filters

Picture Rate Conversion (PRC) has become a mainstream feature in television sets. Over the years, a complete picture rate conversion system has become quite complex. This results in a larger number of heuristics, whose interactions can be hard to predict. In [L. An and A. Heinrich, 2009] a method is proposed to optimize the last part of the PRC-system, the interpolation of pixel data, given a set of motion vectors. It is shown that such a method is capable of handling halo-reduction. It does, however, require a lookup-table which can be fairly large. Also, generating the class-codes still involves heuristics. In this paper an alternative is presented which aims to address these issues. The interpolation is solved as a regression problem leading to a single set of coefficients, while still solving the occlusion problem.