SAR图像几何精校正的OpenMP-GPU协同处理实现

OpenMP-GPU Co-processing for SAR Image Geometrical Correction

几何精校正是合成孔径雷达(SAR)图像处理的一个重要步骤,随着SAR图像时空分辨率的增加,SAR图像几何精校正生产面临着速度上的挑战。提出了一种分块仿射变换结合OpenMP-GPU协同处理的方法。采用分块仿射变换简化图像坐标位置变换,利用OpenMP并行计算提高仿射变换系数的计算速度,同时采用图形处理器(GPU)对图像坐标转换和亮度值重采样操作进行加速处理。实验结果表明,该方法对比传统CPU串行计算的校正方式具有明显的速度优势。

Geometrical correction is an important step of the application of SAR images, with the scale of SAR image data increasing, SAR image geometrical correction is facing the challenge of processing speed. A method based on affine transformation of divided blocks and OpenMP-GPU co-processing is proposed, the affine transforma- tion of divided blocks simplifies the coordinate transformation which is strictly based on SAR imaging theory, and applying OpenMP to speed up the step of computing affine coefficients, in addition, the coordinate transformation and resampling tively improves is accelerated by the parallel computing power of GPU. The results show that this method can effec- the processing speed of SAR image geometrical correction.