论文部分内容阅读
针对毫米波SAR成像中图像模糊并伴有噪声的问题,研究毫米波SAR成像原理,结合并行计算和图像分层理论,提出了一种基于CUDA的毫米波SAR图像增强算法并行化处理和优化方法。该方法以双边滤波作为分层滤波器,并行计算各像素在基本层和细节层上的处理与融合过程,并根据CUDA架构的特性对算法进行优化。仿真结果表明,该实现方法能够有效提高图像对比度,抑制噪声,增强图像细节;同时,在不影响处理效果的前提下,可以达到77左右的加速比,进一步满足系统实时性的要求。
Aimed at the problem of image ambiguity and noise accompanied by millimeter wave SAR imaging, the principle of millimeter wave SAR imaging is studied. Combining with parallel computing and image layering theory, a novel CUDA-based parallel processing and optimization method for SAR image enhancement is proposed . In this method, bilateral filtering is used as a layered filter, and the processing and fusion of each pixel at the basic layer and the detail layer are calculated in parallel. The algorithm is optimized according to the characteristics of the CUDA architecture. The simulation results show that this method can effectively improve the image contrast, suppress the noise and enhance the image detail. At the same time, without affecting the processing effect, the speedup can reach about 77, which can further meet the real-time requirements of the system.