Kalman filter based motion estimation algorithm using energy model

Digital video signal compression is an important requirement for multimedia systems. It can be performed by block-based motion estimation algorithms, which eventuate into acceptable outcomes in both the compression and quality. Already adaptive Kalman filter framework has been applied to motion estimation problem and various autoregressive models have been utilized in it. The main advantages of this approach are its low computational cost and presented sub pixel accuracy. However, they highly depend on the accuracy of their prediction step. In this regard, energy histograms of blocks are going to be served to improve the mentioned accuracy in this paper. Additionally, a new term will be aggregated to the previously presented adaptive variance computing formula to improve its effectiveness on increasing the PSNR. Empirical results indicate the proposed techniques´ benefits over Kalman filter based motion estimation methods. In contrast to the TSS algorithm, this work increases PSNR, in spite of its lesser required computations.