基于GPU的遥感图像配准并行程序设计与存储优化

遥感图像配准是遥感图像应用的一个重要处理步骤.随着遥感图像数据规模与遥感图像配准算法计算复杂度的增大,遥感图像配准面临着处理速度的挑战.最近几年,GPU计算能力得到极大提升,面向通用计算领域得到了快速发展.结合GPU面向通用计算领域的优势与遥感图像配准面临的处理速度问题,研究了GPU加速处理遥感图像配准的算法.选取计算量大计算精度高的基于互信息小波分解配准算法进行GPU并行设计,提出了GPU并行设计模型;同时选取GPU程序常用面向存储级的优化策略应用于遥感图像配准GPU程序,并利用CUDA(compute unified device architecture)编程语言在nVIDIA Tesla M2050GPU上进行了实验.实验结果表明,提出的并行设计模型与面向存储级的优化策略能够很好地适用于遥感图像配准领域,最大加速比达到了19.9倍.研究表明GPU通用计算技术在遥感图像处理领域具有广阔的应用前景.

Parallel Programming Design and Storage Optimization of Remote Sensing Image Registration Based on GPU

Remote sensing image registration is an important processing step of the application of remote sensing image. With the scale of remote sensing image data and complexity of remote sensing image registration algorithm increasing, remote sensing image registration is facing a new challenge on processing speed. In recent years, the power of computing of GPU (graphic process unit) has been greatly improved, which results that the general purpose computing has had a rapid development. In this paper we combine the computing power of GPU for the common computing and the problem of processing speed of remote sensing image registration to study GPU-accelerated processing algorithm. We select a registration algorithm of high accuracy of calculation which is based on mutual information and wavelet decomposition to design the parallel processing, also we propose a parallel model of GPU for the registration algorithm. At the same time we apply the commonly programming optimization strategy of GPU on storage level in remote sensing image registration GPU programming, and we use the CUDA (compute unified device architecture) programming language to realize it based on nVIDIA Tesla M2050 GPU. Experimental results show that the parallel model and optimization strategy of the storage level could be well applied to the field of remote sensing image registration. In our experiment the maximum speedup is up to 19.9X compared with the serial CPU program. This study also shows that the computing technology of GPU has broad application prospects in the field of remote sensing image processing.