SAR Image Despeckling Based On Cauchy Model And Shearlet Transform

Synthetic Aperture Radar(SAR) has so many characteristics, such as all-day, all-weather, high resolution and penetration with important application value, which has been used in the military, topographic surveying, mapping of geology, mineral resources exploration, cartography, agriculture and forestry, marine and water resources and ecology and so on widely. SAR imaging is different from optical imaging mechanism, which will lead to product coherent speckle so that it will reduce the quality of SAR image to affect the subsequent edge detection, image segmentation and object recognition and other image processing. Therefore, it is indispensable issue to be solved to suppress speckle effectively in SAR image processing. In this paper, further research has been discussed about how to suppress speckle effectively in SAR image. The accomplished work as following.(1) The Cauchy distribution model is introduced and the logarithmic cumulant based on second-kind statistics(Mo LC) is proposed for its parameter estimation method. To estimate the unknown parameters of Cauchy distribution that comparing with other commonly used methods for parameter estimation, such as maximum likelihood estimation(ML) and fractional lower order moment estimation method(FLOM). Using the Cauchy distribution to model a high-resolution SAR image. Through experiment of Monte Carlo, we can know that on the estimation accuracy and running time, the method of logarithmic cumulant is superior to others.(2) The theory of Shearlet is introduced and it describes the main characteristics and discretization method. Taking hard threshold despeckling algorithm as example to analyze the relationship between the key parameters of Shearlet and effect of speckle reduction. Through experiments, indicating the parameters of decomposition level, the numbers of the direction and the reasonable threshold value for suppress speckle so that the effect of despeckle reduction is more better.(3) It’s proposed that despeckling method of using Shearlet based on normal inverse Gaussian distribution(NIG) model. Since the SAR image is a multiplicative model so that we should need to logarithmic transformation for it to addictive model for maximum a posteriori. O n the basis of the Shearlet, a contraction function of coefficients in Shearlet domain is derived based on maximum a posteriori criter ion, and provide the method of the parameter estimates of contraction function. Experiments show that the proposed algorithm has good speckle suppression and edge preservation.