An effective and simple Block Matching Algorithm(BMA) based on 'menu vector' search is proposed for motion vector estimation. The proposed BMA is compared with similar existing techniques in the literature and also interfield DPCM.
In this thesis work, several existing BMA's for motion vector estimation are compared with each other in the view of the motion estimation accuracy and hardware feasibility. In the comparison of its accuracy, the entropy of the prediction error signal and the percentage of the predictable picture area are used as performance measures.
And then, a modified algorithm for motion vector estimation is proposed. This algorithm is based on Ninomiya's Recursive Block Matching technique, but its performance is greatly increased by taking into consideration of the weighted summation of the motion vectors in the neighboring blocks as an initial estimate of the motion vector of the subblock in question in the present field.
The performance of the proposed algorithm is 3-4dB better than Ninomiya's algorithm in NMSE(Normalized Mean Square Error).
Especially, the proposed algorithm is an advantageous method in the view of the hardware implementation, because it has only one sequential step for motion vector estimation.
