首页> 外国专利> METHOD OF BLOCK-MATCHING MOTION ESTIMATION WITH FULL SEARCH IN A VIDEO SEQUENCE AND CORRESPONDING ARCHITECTURE

METHOD OF BLOCK-MATCHING MOTION ESTIMATION WITH FULL SEARCH IN A VIDEO SEQUENCE AND CORRESPONDING ARCHITECTURE

机译:视频序列和相应架构中全搜索的块匹配运动估计方法

摘要

Method of motion estimation in a video sequence by means of a block-matching with full search. Firstly the current video frame (1) that forms the sequence is divided into a plurality of reference macro-block (MB), and each macro-block (MB) is divided into a plurality of sub-blocks (SB). Then a search window (3) is chosen in a video frame (2) previous to the current frame (1) and a SAD (Sum of Absolute Difference) is calculated between the pixels of a first reference sub-block (SB) of the current frame (a) and all the ones of the sub-blocks (SB) of equal size (b) in the search window (3). Then the SADmin are determined between all the calculated SAD. Repeating the calculus for each further sub-block (SB) the MV of the macro-block (MB) is computed. An architecture for carrying out this search has two data loading lines (9x, 9y) of reference block (a) and of candidate block (b); a matrix (10a) of processor Element (11) for loading the data of the reference block (a) and comparing them with the data of the candidate block (b); a buffering resource (10b) for adapting the serial input (9y) of the data with their parallel processing (9x, 9y) carried out by the matrix (10a) of the PE (11); an accumulator (20) of the partial sums computed by the matrix (10a) or the PE (11); a Motion Vector processor (30) for computing the MV of reference block (a) with respect to the candidate blocks (b).
机译:通过完全匹配的块匹配在视频序列中进行运动估计的方法。首先,形成序列的当前视频帧(1)被划分为多个参考宏块(MB),并且每个宏块(MB)被划分为多个子块(SB)。然后,在当前帧(1)之前的视频帧(2)中选择搜索窗口(3),并在该帧的第一参考子块(SB)的像素之间计算SAD(绝对差之和)。搜索窗口(3)中当前帧(a)和大小相等的所有子块(b)(b)。然后在所有计算的SAD之间确定SADmin。对每个另外的子块(SB)重复演算,计算出宏块(MB)的MV。用于执行此搜索的体系结构具有参考块(a)和候选块(b)的两条数据加载线(9x,9y);处理器元件(11)的矩阵(10a),用于加载参考块(a)的数据并将它们与候选块(b)的数据进行比较;缓冲资源(10b),用于通过PE(11)的矩阵(10a)执行的并行处理(9x,9y)来适配数据的串行输入(9y);由矩阵(10a)或PE(11)计算出的部分和的累加器(20);运动矢量处理器(30),用于计算参考块(a)相对于候选块(b)的MV。

著录项

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号