Noise reduction methods for terrain phase estimation of InSAR images

Despite decades of scientists' effort, absolute phase determination of interferometry synthetic aperture radar (InSAR) image still remains unsolved. InSAR measurement derived from phase data is not only ambiguous by its modulo-2pi mathematical-ill pose, but also further corrupted by noise. Therefore, we suggest an adaptive least mean square (LMS) algorithm based on steepest descent method for noise reduction purpose. Besides, a scheme which incorporates such filter into a Itoh two-dimensional phase-unwrapping is proposed. The phase estimation procedures are implemented to reconstruct a simulated interferogram of terrain structure. For the quantitative assessment, we employ different types of quality metrics to measure the estimated outcome of InSAR terrain image which includes root mean square error (RMSE) and signal-to-noise ratio (SNR). The estimated outcomes are also reconstructed into three dimensional plotting for visual assessment. By refering to the similar scheme, other noise reduction filters include wieners filter and median filter are implemented for performance comparison. The simulated results show that the proposed method is able to filter noise without corrupted the useful phase information and achieves lowest error energy among other filters. Thus, it is a valuable technique for InSAR terrain phase estimation.