基于GPU的大影像正射校正分块处理方法

Processing Method for Large Scale Image Blocking Orthorectification Based on GPU

图像正射校正处理具有计算复杂度和并行性高的特点,适合在GPU中进行大规模并行处理。然而由于大幅面的影像超过了GPU全局存储器的范围,无法直接导入全局存储器中进行处理。本文提出了影像分块处理策略,对大幅面影像分块然后依次传输到全局存储器中进行处理。实验表明:与CPU串行正射校正算法相比,GPU分块并行处理算法达到了很好的加速效果,加速比达到了188.34倍。

Orthorectification has the characteristics of high computational complexity and high parallelism, and is suitable for large scale parallel processing in GPU. However due to large format image exceeds the scope of the GPU global memory,unable to directly import into the global memory for processing. this paper presents the image block processing strategy, large format image block followed by transfer to the global memory for processing. The experimental results show that compared with the CPU serial correction algorithm, the GPU block parallel processing algorithm achieves a good acceleration effect, and the speedup is up to 188.34 times.