A New Denoising Method of SAR Images
In this paper, we proposed a new approach to separate noise and source signals from SAR images. In the first stage, we used EMD to decompose the SAR image as a collection of some oscillatory basis components termed intrinsic mode functions (IMFs). At the
- PDF / 352,610 Bytes
- 4 Pages / 439.37 x 666.142 pts Page_size
- 29 Downloads / 219 Views
A New Denoising Method of SAR Images Jianguo Zhang
Abstract In this paper, we proposed a new approach to separate noise and source signals from SAR images. In the first stage, we used EMD to decompose the SAR image as a collection of some oscillatory basis components termed intrinsic mode functions (IMFs). At the second stage, PCA is applied to these IMFs to produce uncorrelated and dominant basis components. The most important advantages of our method are as follows: (1) It is not necessary that the components in the SAR images to be linear. (2) Separation process can be performed using only a single mixture. In this paper, we employed the proposed separating model to separate speckle noise from SAR images. Experimental results confirmed the strong potential of the proposed method for speckle noise suppression. Keywords Empirical mode decomposition • Independent component analysis • Synthetic aperture radar image • Speckles removal
61.1 Introduction Various spatial-domain filters have been proposed for the speckle reduction [1, 2]. However, [3, 4], the performance of these filters is heavily dependent on the choice of the size and orientation of the local window. Jain [5] introduces a homomorphic approach, where a SAR image is log-transformed to make the multiplicative speckle noise additive. The log-transformed image is subjected to Wiener filtering, followed by an exponential operation on the filtered output to obtain the despeckled image. However, this method essentially being low-pass filtering blurs many important signal features. In recent years, the multiscale wavelet transform has been used with considerable success for recovering signal from noisy data J. Zhang (*) Department of Foundation Course, Military Economic Academy, Wuhan 430035, China e-mail: [email protected]
Z. Zhong (ed.), Proceedings of the International Conference on Information Engineering and Applications (IEA) 2012, Lecture Notes in Electrical Engineering 217, DOI: 10.1007/978-1-4471-4850-0_61, © Springer-Verlag London 2013
485
486
J. Zhang
[6]. It has been show that wavelet thresholding in a homomorphic framework can provide a better reduction in the speckle noise when compared with that of the spatial-domain filters. However, by employing wavelet shrinkage based on Bayesian formalism, it is possible to recover signals from the noisy data more effectively than by using the thresholding technique [7]. Thus, a number of authors have developed homomorphic methods [8], wherein a suitable probability density function (PDF) is used as a prior model for describing the wavelet coefficients corresponding to the log-transformed SAR image are denoised by means of a Bayesian estimator developed using the prior PDF. Achim have developed a maximum a posteriori (MAP) filter using the symmetric alpha-stable PDF as a prior model. However, the alpha-stable PDF does not have a closed-form noisy data, and increasing the complexity of the Bayesian estimation process. The paper brought forward the particle noise filter algorithms for SAR image.
Data Loading...
