A spatial-temporal de-interlacing algorithm

In this paper, we proposed a spatial-temporal de-interlacing algorithm for conversion of interlaced video to progressive video. Our proposed algorithm estimates the motion trajectory of three consecutive fields interpolates the missing field along the motion trajectory. In the motion estimator, the unidirectional motion estimation and the bidirectional motion estimation processes are combined by multiple objective minimization technique. The unidirectional motion estimation estimates the motion trajectory by comparing the blocks from opposite parity fields while the bi-directional motion estimation compares blocks from the same parity fields. By combining the two motion estimations, the motion trajectory can be accurately predicted. In addition, a quality analyzer is proposed to evaluate the visual quality of the reconstructed frame, which chooses the appropriate interpolation scheme in order to provide maximum de-interlacing performance. Simulation results show the proposed algorithm has better performance over existing de-interlacing algorithm. © 2005 IEEE