Y. Luo and M. Celenk (USA)
Block-matching algorithm, motion vector, motion estimation, spatial correlation, temporal correlation
In this work, we describe a fast block-matching algorithm for motion estimation (ME). The performance of motion estimation is improved by using an adaptive rood search pattern and utilizing the spatio-temporal correlation between the macro-blocks (MBs) for motion prediction. The block-matching is carried out in two stages. In the initial search stage, a novel scheme is proposed that utilizes the temporal and spatial motion correlation between the blocks to choose the candidate neighbouring block for motion prediction for the current block. In addition, based on the available motion vectors (MVs) of the neighbouring MBs, a more flexible strategy is devised which adjusts the lengths of the horizontal and vertical rood arms adaptively and separately. After the initial search, a fixed-size rood pattern is employed repeatedly until the best matching block is found. Experimental results show that the proposed method has better performance than the well known diamond search (DS) and adaptive rood pattern search (ARPS) algorithms in terms of the search speed and the peak-signal-to-noise ratio (PSNR) of the motion compensated images.
Important Links:
Go Back
