CAJ | 학술논문

为了提高视频编码效率，提出一种基于多极小值粒子群的快速运动估计算法.该算法将运动矢量特性和多极小值粒子群算法的全局搜索特性结合，采用自适应运动强度、运动矢量预测以及提前终止迭代等方法，克服单峰误差曲面假设的限制.实验结果表明，对运动平缓和中等的视频序列，该算法的运算复杂度与DS相当.对于运动剧烈的视频序列，该算法的运算复杂度与TSS相当.在增加少量搜索点数的情况下，各类视频序列的搜索精度都接近FS.
위료제고시빈편마효솔，제출일충기우다겁소치입자군적쾌속운동고계산법.해산법장운동시량특성화다겁소치입자군산법적전국수색특성결합，채용자괄응운동강도、운동시량예측이급제전종지질대등방법，극복단봉오차곡면가설적한제.실험결과표명，대운동평완화중등적시빈서렬，해산법적운산복잡도여DS상당.대우운동극렬적시빈서렬，해산법적운산복잡도여TSS상당.재증가소량수색점수적정황하，각류시빈서렬적수색정도도접근FS.
A fast motion estimation algorithm based on multi-miminum particle swarm optimization was proposed to improve the video coding efficiency. By integrating the characteristic of motion vector with the global searching of MMPSO and using strategies like adaptive motion intensity,motion vector prediction and early termination criteria of iteration,the proposed algorithm overcomed the restriction of assumption about the unimodal error curved surface. The experimental results showed that the computational complexity of algorithm was similar to the DS for the video sequences with slow and middle motion. For the video sequences with violent motion,the computational complexity of algorithm was close to the TSS. In the case of increasing a few search points,the search accuracy was close to the FS for all kinds of video sequences.