一种降低硬件HEVC帧间编码器存储器开销的方法  

A Method for Inter Prediction Unit of Hardware HEVC Encoder to Reduce Memory Cost

作　　者：魏家聪 范益波[1] 曾晓洋[1] WEI Jiacong;FAN Yibo;ZENG Xiaoyang(Department of Microelectronics, Fudan University, Shanghai 201203, China)

机构地区：[1]复旦大学微电子学院,上海201203

出　　处：《复旦学报（自然科学版）》2020年第2期194-202,共9页Journal of Fudan University：Natural Science

基　　金：国家自然科学基金面上项目(61674041)。

摘　　要：高分辨率视频中使用了更多的像素来描述空间中相同幅度的运动,这会导致帧间编码过程中需要搜索的范围与所需要的参考帧像素存储器大小随着视频中图像对角线像素数目的平方增长,从而给硬件帧间编码器的设计带来巨大的困难.本文提出了一种以编码树单元为单位,利用当前帧及时域相邻帧中的信息,确定搜索起点的方法;而且在编码效率仅下降约1.6%的情况下,将参考帧像素存储器的规模减小约91.5%.Since more pixels are used in high resolution video to describe motion of the same magnitude in space,it results that the range of inter frame coding needs to be searched,as well as required reference frame pixel memory size increases with the square of the number of pixels diagonal to the image in the video.This introduce tremendous difficulties to the design of the hardware inter-frame encoder.This paper proposes a method to determine the search starting point of a coding tree unit by using the motion information in the adjacent block of the current frame and temporal adjacent frames.In the proposed method,the size of the reference frame pixel memory is reduced by about 91.5%,while the coding efficiency is only decreased by about 1.6%.

关 键 词：高效率视频编码 帧间编码 运动搜索 

分 类 号：TP36[自动化与计算机技术—计算机系统结构] TN37[自动化与计算机技术—计算机科学与技术]