Method for realizing acceleration of anisotropic diffusion filtration of overlarge synthetic aperture radar (SAR) image by graphic processing unit (GPU)

Abstract

The invention discloses a method for realizing the acceleration of anisotropic diffusion filtration of an overlarge synthetic aperture radar (SAR) image by a graphic processing unit (GPU), which solves the problem of low speed caused in the process of processing the overlarge SAR image by adopting anisotropic diffusion filtration. An anisotropic diffusion and filtration process adopts a compute unified device architecture (CUDA) and executes the following steps in the GPU: (1) copying image data I from a host memory of a computer to a memory area A of the GPU; (2) computing diffusion dimension data c(q) of the image data I by using an anisotropic diffusion dimension function; (3) computing the graphical data of an anisotropic diffusion filtration result according to an anisotropic diffusion dimension functional equation; and (4) circularly repeating step (2) and step (3) for T times to obtain the final anisotropic diffusion filtration result graphic IT, and copying the data IT in the memory area C to the host memory of the computer after iteration ends. In the invention, the acceleration is accomplished by the GPU parallel computation under the CUDA, compared with the central processing unit (CPU) serial computation, the processing speed is obviously improved, and the method can be applied to sites with high real-time processing requirements.

Description

technical field [0001] The present invention relates to the field of remote sensing image processing, and more specifically to the field of SAR image filtering. It is a method for accelerating the anisotropic diffusion filtering of ultra-large SAR images by using a GPU, and is used to improve the processing speed of image filtering. Background technique [0002] Speckle noise in radar images greatly reduces the readability of images, which is not conducive to image interpretation and information extraction. In this regard, researchers at home and abroad have done a lot of research work. An ideal filtering method should be able to adaptively smooth the speckle noise, keep the sharp change of the edge and feature boundary, and keep the texture information at the same time. [0003] In the past two decades, image processing methods based on partial differential equations (Partial Differential Equations, PDE) have been greatly developed, and its application range covers almost t...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to overcome the deficiencies of the above-mentioned problems, propose a super large SAR image anisotropic diffusion SRAD filtering

Images