Abstract:

Using the characteristics of translation invariance, multi-resolution of dual tree complex wavelet transform and the more flexible, multi-selectivity of Shearlet transformation, this paper proposes a new algorithm called complex Shearlet transform. A Gaussian mixture model is introduced in order to capture the local coefficients of the complex Shearlets of synthetic aperture radar (SAR) images. The coefficients are estimated by Bayesian least squares estimator based on the model. Then, an inverse complex Shearlet transform is applied to the modified coefficients to get the SAR image after de-noising. The simulation effect and the analysis of the test results show that, compared with other de-noising methods, this algorithm has a better peak signal to noise ratio (PSNR) and the de-noised images are smoother. The computing speed is more than twice as fast as the method using the Shearlet domain Gaussian mixture model de-noising method.

Key words: Shearlet de-nosing, Gaussian mixture model, complex Shearlet transform, SAR image de-nosing, speckle noise