A Novel De-interlacing Technique Using Bi-directional Motion Estimation

A novel de-interlacing algorithm is proposed to accomplish high quality video format conversion. In the proposed scheme, bi-directional motion estimation (ME) is performed between the same parity fields, i.e., the previous and next fields. The bi-directional motion estimation is composed of initial motion vector estimation, motion vector smoothing using motion vector field segmentation, and motion vector refinement in the interpolated field. The proposed method using bi-directional motion estimation can produce robust motion vectors in sequences with camera motion such as panning and zooming. Experimental results show a high visual performance of the proposed de-interlacing algorithm.