计算机工程与应用 ›› 2014, Vol. 50 ›› Issue (15): 18-24.

• 博士论坛 • 上一篇    下一篇

基于UMV的重构图像扩展算法及VLSI实现

朱胜利,姚素英,徐江涛   

  1. 天津大学 电子信息工程学院,天津 300072
  • 出版日期:2014-08-01 发布日期:2014-08-04

VLSI design of reconstructed image padding algorithm for UMV

ZHU Shengli, YAO Suying, XU Jiangtao   

  1. School of Electronic and Information Engineering, Tianjin University, Tianjin 300072, China
  • Online:2014-08-01 Published:2014-08-04

摘要: 在AVS视频解码系统中,针对运动补偿所需的参考宏块出界导致视频解码效率降低的问题,提出一种新型图像边界扩展算法重构了出界参考宏块,以获得相应运动向量进行完善的运动补偿。该算法基于视频解码系统流水线设计要求,同时考虑消除中间存储的需求,对图像数据进行分析,依据图像数据的不同类型和位置进行边界宏块像素的有序扩展,并采用二维存储机制提高效率,使得在运动补偿中实现了图像无限制运动向量(UMV),优化了图像运动补偿效果,提高了视频的解码效率。该算法采用Verilog HDL语言描述,经EDA软件仿真、综合及FPGA验证表明:该算法能够在图像运动补偿中获得出界参考宏块的运动向量,实现图像运动补偿的优化,本模块打开UMV后综合的芯片面积仅增加面积1.5%,而AVS视频解码系统采用UMV比不采用UMV解码效率明显提高。

关键词: 非限制运动向量, 运动补偿, 宏块

Abstract: A novel reconstructed image padding algorithm is proposed for UMV(Unrestricted Motion Vectors) during motion compensation in the AVS video decoding system. Applying this algorithm, outside reference macro blocks are first effectively reconstructed in order. Then motion vectors of outside reference macro blocks can be calculated to optimize performance for motion compensation. VLSI architecture design and implementation in terms of Verilog HDL are also demonstrated. Simulations are carried out by means of EDA tools. Experimental results show that the algorithm can improve the efficiency of motion compensation in a real-time AVS video coding system. In the mean time, a VLSI processor is also designed to correct logic functions and the silicon area of implementation is only added up to 1.5% after UMV is open.

Key words: Unrestricted Motion Vectors(UMV), motion compensation, macro block