Medical Image Sequences Motion Estimation Study

The compression of medical image series is very important in telemedicine. As anessential part of the video frequency compression technology, motion estimation canbe acquired by means of the temporal correlation that exists between the frames ofvideo sequences. Block matching algorithm (BMA) is one of the motion estimationalgorithms.In the existing block-matching algorithms, the full search algorithm (FSA) has thehighest accuracy level and computational complexity. So far, a lot of fast searchalgorithms have been proposed. These algorithms make a trade-off between theaccuracy and the computational complexity. In this paper, a improvedSDS(Square-Diamond Search) algorithm is proposed for the motion estimation ofmedical image series. The improved SDS algorithm reduces the number of thesearched points. This algorithm is used in wavelet transformation field to estimate themotion of medical image series. A simulation experiment for DSA(Digital SubtractionAngiography) is made. The experiment results show that the algorithm accuracy ishigher than that of other algorithms in the motion estimation of medical image series.These fast search algorithms are all based on an assumption, that is, the matchingfunction is monotone. Nevertheless, in practical applications, this assumption cannotcome into existence. So the optimal solution is often substituted by the suboptimalone when these algorithms are used. Based on natural selection theory and naturalgenetic mechanism, genetic algorithm (GA) is broadly used to search optimal solution.But the slow convergent speed and the premature phenomenon make the conventionalGA (CGA) difficult to be used in real time motion estimation. We also proposed aquantum genetic block matching algorithm (QGBMA) by combining the quantumgenetic algorithm with the center-biased motion vector (MV). The experiment resultsshow that the accuracy of QGBMA is higher than three-step search algorithm (TSSA)and is higher than that of diamond-search algorithm (DSA) at large probability. Theeffectiveness of this algorithm is verified.