An Improved Order-Statistical Filter for InSAR Phase Estimation

The task of interferometry phase extraction has been investigated for more than 30 years to reconstruct true phase value from noisy observation. However, it still remains as a typical difficult inverse problem. InSAR observed value is not only ambiguous by its mathematical-ill wrapped pose, but also induced by spatial under sampling issues in areas of phase discontinuities. In addition, phase information is further degraded by inevitable noises. In this letter, an improved order-statistical filter is proposed for InSAR phase estimation. First, a coherence map is generated. Then, proposed order-statistical filter performed its noise filtering based on the coherence map. The filtering rate and its local window size is varying due to coherency: larger window size and more filtering for high coherency area while smaller window size and less filtering for low coherency area. This letter suggests a scheme to associate such filter with Itoh-based phase-unwrapping for principal phase reconstruction. After phase unwrapping, iterative averaging filter is employed for image smoothing purpose. Proposed method is validated by two inteferometric data sets include simulated artificial data and Long’s Peak mountainous data. This method is compared with two other common used denoising filters include Wiener filter and Median filter. Achieved result shows that proposed filter outperform other filters in noise reduction and edge preservation.