基于统一设备架构的影像实时MTF补偿方法  

Optical Image Real-time MTF Compensation Method Based on CUDA

作　　者：孙业超[1] 朱晓波[1] 郝雪涛[1] 

机构地区：[1]中国资源卫星应用中心,北京100094

出　　处：《航天返回与遥感》2013年第4期93-99,共7页Spacecraft Recovery & Remote Sensing

摘　　要：调制传递函数补偿(Modulation Transfer Function Compensation,MTFC)处理可以恢复光学影像在成像、传输和处理过程中损失的高频信息,使得影像边缘更为清晰、细节更为丰富,从而改善影像质量。然而,随着遥感对地观测技术的不断发展,高空间分辨率影像数据量大,导致复原处理时计算量大、耗时长,成为高分辨率地面处理系统实时处理的瓶颈问题。为解决这一问题,提出了基于统一设备架构模型的MTFC并行处理方法,并根据算法特点对存储结构和并行计算结构进行了优化改进,利用真实卫星影像进行了验证实验。实验结果表明,在保证影像复原质量的前提下,提出的算法获得了45倍的加速比,可以满足地面系统实时处理的要求。The modulation transfer function compensation （MTFC） processing can restore the lost high-frequency information of the optical image in the course of imaging, transmission and processing, making images clearer and richer detail edge, thereby improving image quality. However, owing to the large volume of data of high-resolution images, the image restoration generates huge computational load and is time-consuming. To solve this problem, this paper proposes and implements a parallel MTFC algorithm based on the compute unified device architecture （CUDA） model. The experimental results show that under the premise that the qual- ity of image restoration is assured, the speedup can reach 45 times, and processing efficiency can be promoted greatly.

关 键 词：统一设备架构光学 调制传递函数补偿 高分辨率 遥感数据 地面处理系统 

分 类 号：TN75[电子电信—电路与系统]