Successive elimination motion estimation algorithm based on multi-resolution

This paper proposes a new motion estimation method based on the adaptive break successive elimination algorithm and multi-resolution in a digital image stabilization (DIS) system. A computational scheme that facilitates the estimation of the local motion vectors is developed. First, the images are transferred to layers with different resolutions by the linear or wavelet transition and the performances of the translations are compared in the later experiments. With multi-resolution, the size of the sub-image and the search area were broadened and little moving objects can be eliminated. Then, positions which were calculated with lower possibility to be the destination by successive elimination algorithm were removed before they are matched. Finally, also with the adoption of the principle of adaptive break, the threshold which decides the times of a block matching is set adaptively according to the errors during the match while the accuracy remaines. Experimental results show that with similar performance as that of TSS, the computational load of the proposed motion estimation algorithm is 3% of the full-search matching and 34% of TSS..