Adaptive matching scan algorithm based on gradient magnitude for fast full search in motion estimation

To reduce an amount of computation of the full search algorithm in motion estimation for video coding, we propose a new and fast matching algorithm without any degradation of predicted images compared with the conventional full search algorithm. In this paper, we show that the block matching error is proportional to the image complexity of the reference block in the current frame. The computational reduction results from an adaptive matching scan algorithm according to the gradient magnitude of the reference block. Experimentally, we significantly reduce the computational load, while keeping same error performance as that of the full search.