A Fast Speckle Reduction Algorithm Based on GPU for Synthetic Aperture Sonar

Abstract

Synthetic aperture sonar (SAS) is a kind of high resolution imaging sonar, but speckle exists in SAS image for that SAS is a coherent imaging system, which makes it very difficult to visually and automatically interpret. In this paper, a fast speckle reduction algorithm for SAS is proposed in GPU environment, which helps to solve the speckle reduction problem of SAS image in real-time. Firstly, the original SAS image is partitioned into small rectangular blocks with partly overlapped and uploaded into the shared memory of GPU by block of threads. Secondly, the Lee filtering process is performed on every blocked SAS image by the multi-core of GPU simultaneously. Finally, the whole processed result is obtained by merging those local filtered images. The feasibility and high efficiency of the method are confirmed by the speckle reduction experiment on the real SAS image.