一种快速合成孔径声纳图像相干斑抑制方法

摘要

A fast speckle reduction algorithm for synthetic aperture sonar (SAS ) is proposed in GPU environment , which helps to solve the low efficiency of speckle reduction of SAS image .Firstly ,the whole SAS image is uploaded to the memory of GPU ,and partitioned into small rectangular blocks with partly overlapped according to the size of the local win -dow .Secondly ,the Lee filtering process is performed on every blocked SAS image by block of threads simultaneously .Final-ly ,the whole processed result is obtained by merging those local filtered images and downloaded to the memory of the host . The feasibility and efficiency of the proposed method are verified through experiment of real SAS image speckle reduction .%针对合成孔径声纳图像相干斑抑制效率低的问题，提出了一种 GPU 异构环境下的合成孔径声纳图像相干斑抑制方法。首先将合成孔径声纳图像上传至 GPU 存储器，然后根据滤波窗口大小将图像分割为重叠小块后采用多个线程块同时进行 Lee 滤波处理，最后对处理结果进行拼接获得最终结果并下载到主机内存。对真实合成孔径声纳图像相干斑抑制试验验证了所提方法的可行性和高效性。