首先简要介绍了统一设备架构CUDA(Compute Unified Device Architecture)技术的背景、特点、内存模型,利用通用计算图形处理单元GPGPU(General Purpose GPU)及CUDA技术,实现了图像直方图均衡化和薄云去除的并行化处理,与传统的基于CPU的方法相比,两个基于GPGPU的图像预处理操作的执行效率分别提高了40倍与80倍左右,在大规模实时性图像处理操作中,有很大的实用价值。
Parallel processing of histogram equalization and removal of light cloud for gray scale images are implemented by GPGPU and CUDA technique. The efficiencies of two image pre-processing operations simulated on GPGPU can be improved up to 40 and 80 times compared to the sequential methods based on CPU, respectively. This approach can be redounded to real-time applications for large scale image processing and analysis.