A Hybrid Motion Compensation De-interlacing Approach

Motion compensation de-interlacing is expected to be better than linear techniques; but all the block-based motion compensation de-interlacing methods cause block artifacts. The algorithm proposed in this paper is concerned with reducing the deficiency of motion-compensated interpolation by using adaptive hybrid de-interlacing methods. A spatio-temporal tensor-based approach is used to get more accurate motion field for de-interlacing. Motion vector is assigned for each position with pixel precision; the block artifact is reduced significantly. To deal with the artifacts introduced by motion-compensation when the motion estimation is incorrect, linear techniques are considered by adaptive weighting. Furthermore, directional filter is adapted to preserve details and the edge discontinuity could be eliminated greatly. Our approach is robust to incorrect motion vector estimation.