Study On Autofocus Technique And Speckle Reduction In SAR Image Formation

Based on the Doppler effect and pulse coherence technique, Synthetic Aperture Radar (SAR) breaks through the azimuth resolution limitation imposed by real aperture antenna. In combination with the pulse compression technique, two-dimensional high resolution imagery to distant targets can be realized. This dissertation studies image formation algorithm, autofocusing technique and speckle reduction method on SAR signal processing. In order to make a comparison between the different processing algorithms, almost all methods are provided with the processing results of real data and simulated data.Chapter 1 introduces firstly the basic principle of SAR imaging and discusses two typical SAR image formation algorithms. Then, the characteristics, examples of application, and developing history of SAR imaging technique are reviewed. Current status of foreign and domestic SAR techniques is summarized and main trend of airborne SAR is predicted. Finally, this chapter emphasizes the importance of auxiliary processing in SAR imaging, indicates that autofocusing is essentially a problem of blind deconvolution, and that speckle reduction is a problem of imagery restoration. No additional assumption and limitation, the inverse problem on autofocusing or speckle reduction can not be solution.Chapter 2 introduces the sources and influences on images of all kinds of phase errors, analyzes existing some autofocusing algorithms and lays stress on considerations for application. Since Phase Gradient Autofocusing (PGA) often produces unsatisfactory results when operating upon scenes that are dominated by distributed targets, a modified Weighting Average phase errors estimation method is proposed. The processing results of real data show that this method reduces effectively iterative degree and gets visual effect that is analogous to that processed by PGA. On the basis of Contrast Maximization Autofocusing algorithm (CMA), a method based on Minimum Entropy Criteria (MEC) for SAR imagery autofocusing is proposed. Simulated results show that entropy is more suitable for a measure of degree of image focus than contrast. The processing results of real data show that the effects processed by CMA and MEC is equivalent. On the assumption that imagery pixel follows non-normal distribution in high resolution imaging, a new method of phase errors estimation based on higher-order statistics is proposed for SAR imagery autofocusing. Experimental results show the feasibility of the proposed method. Lastly, the dissertation makes a rough comparison between all autofocusing algorithms and points out that the performance of autofocusing is dependent on particular scenes and the characteristics of phase errors.With the inherent relation between spotlight mode SAR and stripmap mode SAR, a method that stripmap SAR imaging is performed by multiple-subpatch processing algorithm based on spotlight mode SAR is also proposed in Chapter 2. In the method, we study how to choose scene size to avoid significant motion through resolution cells and develop the formula of filter bandwidth determined by scene size. The feasibility of the proposed method is verified by real data. Presented is a solution of PGA algorithm for stripmap SAR autofocusing via data conversion. Its validity is proved by the processing results of real data.Chapter 3 makes a comprehensive research into statistical properties of speckle noise. Multilook processing and some space-domain filtering methods for speckle reduction are discussed. In the light of wavelet-domain soft-thresholding denoising algorithm, with hidden Markov models (HMMs) structure of images, a new speckle suppression method based on wavelet-domain statistical filtering is proposed. Simulation and experimental results using real data are applied to quantitatively and qualitatively compare the proposed method with some typical filtering algorithms. The conclusion is that the wavelet-domain HMMs method effectively improves equivalent number of look, preserves spatial resolution of raw imagery. In addition, the pro...