摘要
针对多模式的块匹配运动估计技术中的运动搜索算法部分,研究了目前广泛使用的如三步法、菱形法等经典搜索算法,分析了其在实现编码过程中性能上速度和精度两方面的优势与劣势。在此基础上,结合相邻宏块的运动相关性,通过预测MV判断搜索起始点,并且在运动搜索中引入了一种依据阀值的提前终止策略,从而提出了一种新的带有阀值的自适应混合搜索算法。通过采用经典的JM8.6模型进行的仿真验证,结果表明该算法既显著降低了编码复杂度,又保证了预测的有效行和健壮性,具有很强的实用性。
For the search algorithm of multi-mode block-matching motion estimation, the most popular search algorithms, such as three-step method, diamond method and other classic algorithms, are studied. Meanwhile, the advantages and disadvantages of encoding speed and accuracy of them are also analyzed. Based on this, combined with movement correlation of adjacent macro blocks, the starting point is searched by predicting MV judge, and an early termination policy based on threshold values is introduced into sports search, thus a new adaptive hybrid search algorithm with threshold values is put forward. The simulation by classic JM8.6 model proves that, this algorithm could simplify the encoding, and ensure the efficiency and robustness. It is of great practicability.
出处
《计算机工程与设计》
CSCD
北大核心
2010年第9期1968-1970,共3页
Computer Engineering and Design
