A modified joint-pixel based SAR interferogram auto-registration and denoising method

Abstract

A modified joint-pixel algorithm for autoregistration and interferometric phase denoising is proposed in this paper. For the blindness of the sample selection in joint-pixel method, this paper first analyzes the distribution that the sample obeyed. Under the guidance of this distribution, amplitude and phase estimation models of joint-pixel method are established. Through the preprocessing on the phase and amplitude of the two interferometric SAR images, an effective sample to estimate the true value is achieved. Moreover, taking advantage of the coherence information of the effective samples, the proposed method automatically registers the SAR images while denoising without the loss of detail. Compared with the original method, the interferogram can be estimated more accurate and the influence of the abnormal amplitude points on the surrounding is reduced. Furthermore, in order to solve the contradiction of preserving texture and filtering strength, an iterative algorithm is invented. In the end, the effectiveness of this modified algorithm is validated by both simulated and real data.