Speckle Reducing Anisotropic Diffusion Based On Modified Diffusion Coefficient

Synthetic Aperture Radar has been widely used in many fields due to its advantages such as its all-time and all-weather properties. The special way of imaging, in SAR image, makes it to have a serious impact on the understanding and interpretation which is caused by speckle, so how to reduce speckle has been an urgent problem.In this paper, three kinds of modified Speckle Reducing Anisotropic Diffusion(SRAD) algorithms are respectively proposed.The main works of this paper are as follows:(1) A modified SRAD algorithm based on difference curvature (DCSRAD) is given. Since the traditional SRAD algorithm can not obtain enough smooth performance in uniform regions, a modified diffusion coefficient, which uses an edge operator by combining the difference curvature with the instantaneous coefficient of variation is provided. According to theoretic analysis, we can verify that DCSRAD algorithm is isotropic in homogeneous so that speckle can be reduced effectively. Experiments show that DCSRAD algorithm has a good performance.(2) A modified SRAD algorithm based on conditional information(CISRAD) is proposed. Instantaneous coefficient of variation is used to construct the diffusion coefficient in the original SRAD algorithm. The modified diffusion coefficient based on conditional information is proposed which satisfies Charbonnier’s Criterion in both homogeneous regions and heterogeneous regions. The modified diffusion coeffient is well-posed so that it can make the iteration stable, furthermore it is not sensitive to noise. Experiments indicate that CISRAD algorithm has better results in speckle reducing and structure preserving.(3) A modified SRAD algorithm based on Gaussian-Gamma bi-windows (GGSRAD) is presented. In SAR image, the probability of the current pixel which is an edge or not can be measured by using Gaussian-Gamma bi-windows, so the modified diffusion utilizing Gaussian-Gamma bi-windows is proposed. According to the theoretical analysis, the modified diffusion coefficient satisfies the Charbonnier’s Criterion not only homogeneous regions but also heterogeneous regions. In heterogeneous regions, GGSRAD algorithm can improve the ability to preserve structure. The interation is stable because the modified diffusion coefficient is well-posed. Experiments demonstrate GGSRAD algorithm has good ability in speckle reducing and structure preserving ability.