论文部分内容阅读
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。
In order to improve the processing efficiency and speed of remote sensing images, the corresponding image restoration and super-resolution algorithms and their parallel hardware architecture are studied. First of all, the key algorithms such as defuzzification, denoising and super-resolution in the frequency domain are introduced. Secondly, the architecture of DSP and cluster are analyzed and compared. The results show that the cluster is more suitable for large-scale parallel processing. Finally, DSP and computer system on the processing efficiency and speed of the algorithm were tested and analyzed, given the performance prediction of the cluster, to determine the key parameters that the choice of the number of processors. Experimental results show that the system can meet the requirements of real-time applications, while the sharpness, contrast and resolution of the processed images are significantly improved.