Research On SAR Images Despeckling Algorithm Based On Sparse Representation

Synthetic aperture radar(SAR) images have been widely used in plenty of fields of national security and economy. However, the SAR images contain speckle because of the coherence effects between the backscattered signals when imaging. The speckle seriously affects the SAR images quality. In order to understand and interpret the SAR images, it is necessary to despeckle but keep the details at the same time.The main work of this paper is on the dictionary learning algorithm and the image fusion technology which is applied into SAR images despeckling. The multi-dictionary fusion algorithm for despeckling in the SAR images is proposed in this paper, and demonstrate its effectiveness from theoretical and experimental aspects. The main contributions of this paper include the following three aspects:(1)The complexity of the scatter targets make the SAR image doesn’t have an accurate model, so the traditional denoising methods don’t work effectively. However,the appearance of multi-geometric analysis provides a new way to despeckle effectively.Contourlet transform has multi-scale features and multi-directions characteristics, can make the image of two or higher dimensions get more effective geometry representation.As the bandpass image after the Laplacian pyramid filter and bandpass filter has oscillation phenomenon near the singular point in image decomposition, the despeckling of Contourlet transform effect is affected. So this paper further studies the translation invariant Contourlet transform, namely TICT. Experiments show that this method can effectively despeckle, and eliminate pseudo-Gibbs phenomenon up to a certain extent.(2)AS the dictionary learning algorithm can maintain the detail of image more effectively during the denoising process, compared with other traditional denoising methods. This paper introduces the principle of sparse representation and dictionary learning algorithm in detail. On this basis, this paper study the K-SVD(Singular Value Decomposition) and nonparmetric Bayesian dictionary learning algorithm in the application of SAR images despeckling. The experimental results show that the two types of dictionary learning algorithm can not only despeckle effectively, but also can keep the edges and other details of SAR image well.(3)After having a detailed research about Contourlet transform and dictionary learning algorithm in the application of SAR images despeckling, this paper proposed akind of multi-dictionary fusion of SAR images despeckling algorithm. This paper first introduces the application of image fusion technology in image denoising, then choose some image fusion methods to process the images researched by dictionary learning algorithms experimented in this paper. The results of experiments show this method can achieve a higher PSNR( peak signal-to-noise ratio), and retain more edges and texture information.