CAJ | 학술논문

　　随着整像素运动估计快速算法的发展，分像素运动估计的计算量在运动估计中所占比重越发明显。为了减少分像素运动估计的计算量，提出了一种利用运动矢量空间相关性来预测整像素运动块，对整像素运动块进行分像素搜索过程跳过的分像素运动估计方法。实验结果表明，该算法与全分像素搜索算法结合使用，在基本保持搜索精度不变的情况下，比单纯的全分像素搜索算法减少60%左右的分像素搜索点。该算法可与其他快速分像素搜索算法结合使用，以获得更好的编码性能。
　　수착정상소운동고계쾌속산법적발전，분상소운동고계적계산량재운동고계중소점비중월발명현。위료감소분상소운동고계적계산량，제출료일충이용운동시량공간상관성래예측정상소운동괴，대정상소운동괴진행분상소수색과정도과적분상소운동고계방법。실험결과표명，해산법여전분상소수색산법결합사용，재기본보지수색정도불변적정황하，비단순적전분상소수색산법감소60%좌우적분상소수색점。해산법가여기타쾌속분상소수색산법결합사용，이획득경호적편마성능。
With the development of fast integer pixel motion estimation algorithms, the proportion of the computational cost of fractional pixel motion estimation in that of the whole motion estimation has become more and more significant. To reduce the complexity of fractional pixel motion estimation, a novel method of H.264 fractional pixel motion estimation based on the spatial correlation of motion vector is proposed. It predicts Integer Pixel Motion Block(IPMB)using the spatial correlation of motion vector. It skips the fractional pixel search process for IPMB. Experimental results show that combined with the full fractional pixel search algorithm, the proposed algorithm can reduce 60%of fractional pixel search points while maintaining the search accu-racy basically compared with the pure full fractional pixel search algorithm. The proposed algorithm can be combined with other fractional pixel search algorithm to get better coding performance.