基于阵列处理器的SVDC算法并行设计与实现  

PARALLEL DESIGN AND IMPLEMENTATION OF SVDC ALGORITHM BASED ON ARRAY PROCESSOR

作　　者：蒋林 黎瑞金 曹非 Jiang Lin;Li Ruijin;Cao Fei(School of Computer Science and Technology,Xi’an University of Science and Technology,Xi’an 710600,Shaanxi,China;School of Electrical and Control Engineering,Xi’an University of Science and Technology,Xi’an 710600,Shaanxi,China)

机构地区：[1]西安科技大学计算机科学与技术学院,陕西西安710600 [2]西安科技大学电气与控制工程学院,陕西西安710600

出　　处：《计算机应用与软件》2023年第10期285-290,共6页Computer Applications and Software

基　　金：国家自然科学基金重点项目(61834005/F0402)。

摘　　要：视点合成失真算法(Synthesized View Distortion Change,SVDC)作为三维高效视频编码(3D High Efficiency Video Coding,3D-HEVC)中改善深度图编码效率的有效途径,已成为当下三维视频领域的研究前沿之一。基于阵列处理器,利用分布式共享存储结构设计并实现一种SVDC算法的并行映射方式,并根据访存特性提出失真值计算优化方案,以像素级误差平方和(Sum of Squared Differences,SSD)计算替代单元级SSD计算。实验表明,相比于HTM平台,算法的平均性能可以提升19.03%,所设计的失真值计算并行方案串/并加速比为2.36,使用像素级SSD计算后相比于优化前平均性能可以提升39.3%。Synthesized view distortion change(SVDC),as an effective way to improve the efficiency of depth map encoding in 3D high efficiency video coding,has become one of the frontier technologies in the field.This paper designed and achieved a parallel mapping method of the SVDC algorithm by using the distributed shared memory structure,optimized the distortion change calculation based on the memory characteristics,and replaced the unit-level SSD calculation with pixel-level SSD calculation.The experiments show that compared with the HTM platform,the average performance of this algorithm is improved by 19.03%,and the designed distortion value calculation parallel scheme has a serial/parallel acceleration ratio of 2.36.The average performance of the pixel-level SSD calculation is improved by 39.3%compared with former optimization.

关 键 词：阵列处理器 三维高效视频编码 分布式共享存储 视点合成失真算法 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]