摘要
几何校正是分析和应用航空图像的一项重要前期处理工作,工程中存在校正算法实现的实时性问题。在基于多协议交换总线(VerstatileProtocolSwitch,VPX)的通用图像处理器(GPGPU)平台上对图像校正算法进行并行设计,实现了图像校正算法的计算加速。编程对比了几种常用分辨率的航空图像在GPGPU和CPU这2种环境下校正处理的处理时间,处理结果表明,基于GPGPU并行图像校正相比单纯使用CPU方法,其计算加速比可达10~20,较大地提高了处理实时性。
As an important pre-processing in the analysis and application of aviation image,the geometric correction is difficult for real-time processing. Based on the platform of NVIDIA GPGPU and VPX bus,the parallel image correction method is designed to solve the inefficient image transforms. In GPGPU environment and CPU environment, the algorithm is simulated for aerial image of typical size. The experimental results show that the average speedup ratio of GPGPU-based parallel image correction reaches 10 ~ 20 ,compared with the method which only uses CPU. The real-time performance of image correction system is improved.
出处
《无线电工程》
2014年第1期53-55,共3页
Radio Engineering
基金
国家部委基金资助项目
关键词
并行计算
几何校正
坐标系转换
统一计算设备架构
parallel process
geometric correction
coordinate transformation
Compute Unified Device Architecture(CUDA)