FAST BLOCK MOTION ESTIMATION ALGORITHM BASED ON SORTING OF PREDICTION VECTORS

In this paper we present a new fast motion estimation algorithm using spatial and temporal correlation among motion vectors. Motion vector of a block can be predicted from the motion vectors of its neighboring blocks. By statistical analysis we show that the role of different neighboring blocks in the prediction process changes with time. We also show that the number of prediction vectors affect on the matching error. Based on matching errors produced by prediction vectors we suggest a dynamic routine. In our designated algorithm there are some parameters for controlling quality and speed of the algorithm. By selecting these parameters according to our demands (i.e. quality of reconstructed image or speed of the algorithm) we can develop some different algorithms. We show the result of one case in this paper that chooses only one prediction vector produce minimum error. Then we search around this vector to find the best matched block. Experimental results show notable PSNR improvement as well as speed up ratio over well known algorithms

Keywords:

Motion estimation , adaptive algorithm , temporal correlation , spatial correlation