摘要
为了解决天文图像相减测光存在的性能问题,满足特殊条件下天文观测的实时性要求,在充分分析原始测光算法整体性能的基础上,结合CUDA并行编程模型,并行优化测光算法中模板图像降晰的计算,提出并实现了一种GPU加速的测光算法GAISP。实验结果表明,GAISP在处理较大规模天文图像时,图像相减部分耗时较原始算法降低2/3。
To address performance problem about astronomical image subtraction photometry and meet the astronomical observation real-time requirements under the special conditions,this paper analyzed the overall performance of the original photometry algorithm and optimized the template image processing part of the photometry algorithm with CUDA programming model.It designed and implemented new photometry algorithm based on GPU named GAISP.The experimental result shows that the image subtraction part of GAISP increase average performance of 2/3 compared with traditional photometry algorithm when processing the large scale image.
出处
《计算机应用研究》
CSCD
北大核心
2011年第10期3940-3943,共4页
Application Research of Computers
基金
国家自然科学基金资助项目(10978016
11003027)
天津市科技支撑重点项目(09ZCKFGX00400)
关键词
图形处理器
相减测光
并行化
性能优化
GPU
subtraction photometry
parallelization
performance optimization