基于OpenCL的SAR影像快速浏览方法研究  

Research on the Fast Browsing Method of the SAR Images Based on OpenCL

作　　者：孙祥杰 朱亮 余同欢 SUN Xiangjie;ZHU Liang;YU Tonghuan(Nanjing Research Institute of Electronic Engineering,Nanjing 210000,China)

机构地区：[1]南京电子工程研究所,江苏南京210000

出　　处：《电子质量》2023年第3期24-30,共7页Electronics Quality

摘　　要：在分析了SAR影像数据的结构特点、OpenCL的计算架构和影像缩放算法的适用方式的基础上,提出了基于OpenCL的SAR影像快速浏览方法。使用OpenCL异构并行技术加速最邻近插值、双线性插值和双三次插值3种常用的影像缩放算法,并进行全方位的测试评估实验。研究结果显示系统在OpenCL优化的最优配置下能够显著地提高系统内存应用效率和加快SAR影像缩放过程中的加载速度,大幅度地提升影像缩放算法的效率,其实时性和影像表现可以完全满足对于大数据量SAR影像的浏览要求。Based on the analysis of the structure characteristics of SAR image data,the computational architecture of OpenCL and the applicable way of image scaling algorithm,a quick SAR image browsing method based on OpenCL is proposed.OpenCL heterogeneous parallel technique is used to accelerate three commonly used image scaling algorithms,which are nearest-adjacent interpolation,bilinear interpolation and bicubic interpolation,and a comprehensive test and evaluation experiment is carried out.The research results show that under the optimal configuration of OpenCL optimization,the system can significantly improve the application efficiency of system memory and speed up the loading speed in the process of SAR image scaling,and greatly improve the efficiency of image scaling algorithm.Its real-time performance and image performance can fully meet the browsing requirements of large data volume SAR images.

关 键 词：合成孔径雷达 海量数据 OPENCL 异构并行 快速浏览方法 

分 类 号：TP399[自动化与计算机技术—计算机应用技术] TP751[自动化与计算机技术—计算机科学与技术]