遥感图像复原与超分辨算法及其并行实现  

Image restoration and super-resolution algorithms based on frequency domain and their parallel implementation

作　　者：马冬冬[1] 李金宗[1] 朱兵[1] 穆立胜[1] 

机构地区：[1]哈尔滨工业大学图像信息技术与工程研究所,黑龙江哈尔滨150001

出　　处：《光电子．激光》2009年第10期1365-1370,共6页Journal of Optoelectronics·Laser

基　　金："十一五"总装重点资助项目(51322020703)

摘　　要：为了提高海量遥感图像的处理效果和速度,进行了相应的图像复原与超分辨算法及其并行硬件体系结构的研究。首先,介绍了频域内的解模糊、去噪和超分辨等关键算法;然后,对DSP和机群两种体系结构进行了分析和比较,结果表明,机群更适合于大规模并行处理;最后,在DSP和计算机系统上对算法的处理效果和速度进行了实验和分析,给出了机群的性能预测,确定了关键参数即处理器数目的选择依据。实验结果表明,该系统可满足实时应用的要求,同时使处理后图像的清晰度、对比度和分辨率显著提高。Image restoration and super-resolution algorithms with parallel hardware architectures are studied to improve the processing effect and speed of vast remote sensing images.First,the key algorithms of deblur,denoise,super-resolution based on frequency domain are presented.After analyzing and comparing the DSP and cluster,it is concluded that cluster is more suit for large scale parallel processing.Finally,experiments and analyses are made on processing effect and speed of DSP and computer system,to achieve t...

关 键 词：图像复原 超分辨 DSP 机群 并行实现 实时处理 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]