首页> 外国专利> A REAL-TIME HIGH SPEED FULL SEARCH BLOCK MATCHING MOTION ESTIMATION PROCESSOR

A REAL-TIME HIGH SPEED FULL SEARCH BLOCK MATCHING MOTION ESTIMATION PROCESSOR

机译:实时高速全搜索块匹配运动估计处理器

摘要

In the hierarchical block search, the present invention improves the utilization of hardware by eliminating the delay time due to the dependency between several motion estimation layers, and has the scalability and variability based on the unified architecture. The present invention relates to a calculation method for fast real-time processing motion estimation and an apparatus for the same, which can be extended and applied to a device. Pixel motion estimation processor 104, past frame storage memory 105-111 for storing data necessary for operation on one macro block and current frame macro block data 112-118 corresponding to the macro block of each past frame. Is done. The first, second, and third hierarchical search motion estimation processors 101, 102, and 103 perform a motion estimation operation of each layer in parallel, and the past frame data of the processor is based on a motion vector of a higher layer. The frame search area is determined. Thereafter, a delay occurs because the operation must be started only after past frame data of the motion vector is input to the lower layer motion estimation processor. The motion estimation processor 101-104 calculates the past frame macroblocks stored in the past frame storage memories 105, 107, 109, and 111, respectively, based on the unified architecture in various motion estimation calculation modes. It is a very useful invention that can realize high speed operation and increase the redundancy of input data, thereby minimizing the number of input / output pins and minimizing the amount of hardware.
机译:在分层块搜索中,本发明通过消除由于几个运动估计层之间的依赖性而引起的延迟时间来提高硬件的利用率,并且基于统一架构具有可伸缩性和可变性。本发明涉及一种用于快速实时处理运动估计的计算方法及其装置,其可以扩展并应用于设备。像素运动估计处理器104,过去帧存储器105-111,用于存储对一个宏块进行操作所需的数据,以及与每个过去帧的宏块相对应的当前帧宏块数据112-118。已经完成了。第一,第二和第三分层搜索运动估计处理器101、102和103并行地执行每个层的运动估计操作,并且处理器的过去帧数据基于较高层的运动向量。确定帧搜索区域。此后,由于仅在将运动矢量的过去帧数据输入到下层运动估计处理器之后才必须开始操作,因此发生延迟。运动估计处理器101-104基于各种运动估计计算模式中的统一架构,分别计算存储在过去帧存储存储器105、107、109和111中的过去帧宏块。这是一个非常有用的发明,可以实现高速操作并增加输入数据的冗余度,从而最大程度地减少输入/输出引脚的数量并减少硬件的数量。

著录项

  • 公开/公告号KR100246033B1

    专利类型

  • 公开/公告日2000-03-02

    原文格式PDF

  • 申请/专利权人 YOU JAE-HEE;

    申请/专利号KR19970030135

  • 发明设计人 유재희;

    申请日1997-06-25

  • 分类号G06F7/50;

  • 国家 KR

  • 入库时间 2022-08-22 01:45:00

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