Research And Application Of Speckle Reduction Algorithm For Synthetic Aperture Radar Images

The presence of speckle noise in the synthetic aperture radar (SAR) images severely restricts the application of the coherent images. Speckle noise is generated by the coherent processing of radar signals. During the past two decades, many speckle reduction techniques have been developed for removing speckle. An ideal algorithm should smooth the speckle without blurred edges and fine detail. But most algorithms cannot satisfy these two demands very well.Most of the standard algorithms use a defined filter window to estimate the local statistical properties of SAR image. Generally, it could reduce speckle level in homogeneous areas. But the images is either blurred or over smoothed due to losses in detail in non-homogenous areas like edges or lines, so that there solution of the images is reduced, some subtle details are hidden and understanding the images becomes very hard.For reducing speckle noise from Synthetic Aperture Radar (SAR), we describe the basic purpose and significance of speckle smoothing algorithm for SAR images. The history of the speckle reduction is introduced briefly. And also give a presentation of the ground conditions of the areas imaged. Based on principle of coherent speckle noise being generated, we described how the speckle formed in detail. Its statistic mathematic model is be analyzed and its probability-density function; statistics of attribute, mathematics of attribute are given. These are the fundamental properties of speckle in SAR images. To compare speckle smoothing algorithms quantitatively, a set of performance criteria is discussed.The primary coverage of this article can separate into three parts. The contributions of each part are:Realize and analyze several well-known speckle filtering algorithms, such as lee filtering algorithms, Enhanced Lee filtering algorithms, wavelet soft threshold filtering algorithms, maximum a posteriori Probability filtering algorithms, etc. The concept of each filtering algorithm is discussed in detail.This paper describes MAP (Maximum a Posteriori Probability) filtering in detail. After that a new method called Edge_map filter is proposed. The experiment results show that the image edges are retained with the method and the speckle is removed effectively. Finally we discuss this new speckle smoothing algorithms quantitatively.A new method with isotropic diffusion algorithm which is based on anisotropic diffusion equation for MSTAR speckle reduction is proposed. With the local statistics values, it provides anisotropic diffusion in the heterogeneous area and fast isotropic diffusion in the homogeneous area. Thus, it can get the favorite target area image quickly. After image preprocessing, we perform image segmentation by means of self-organizing feature map neural network. In our experiment, this method proved good performance in reducing speckle noise in MSTAR images and work excellent for image segmentation.