Feature-bit-plane motion estimation using enhanced motion vector

This paper presents a fast block-based motion estimation algorithm based on a bit plane technique. This algorithm can produce the eligible candidates for finding the enhanced motion vectors. The proposed method translates the expensive 2D block-matching problem to a simpler 1D matching by choosing some blocks as eligible candidates and eliminating the others. The bit-plane idea combined with the enhanced motion vector leads to a novel motion estimation algorithm. This novel algorithm offers computational scalability through two parameters, so the speed/performance trade-off of the proposed method can easily be controlled. Experiments show that the proposed method is several times faster than the full search algorithm and it produces a better prediction performance. There are some cases in which existing motion estimation techniques obtain a false motion vector with a huge prediction error, while the proposed method can find an enhanced motion vector with an excellent accuracy.