Frame Rate Up-Conversion Using Hierarchical Motion Estimation and Variance-Based Motion Vector Smoothing
In this paper, a new motion-compensated frame rate up-conversion (MC-FRUC) method is proposed. The
proposed algorithm decomposes each frame to a hierarchical structure and estimates motions between two adjacent frames at
the top level using unidirectional motion estimation. These motion vectors (MVs) are refined with motion estimation (ME)
of small search range in the lower levels. False MVs; especially in the edge area, are detected using the variance of residual
values and corrected by weighted sum of neighboring MVs. These MVsare assigned to the block grid of interpolated frame
to avoid holes and overlapped regions. As a result, average peak signal-to noise ratio and structural similarity were increased
up to 2.9 dB and 0.015, respectively compared with other methods, and computation time was reduced.
Index terms- Frame rate up-conversion, motion estimation, motion vector smoothing.