Quality Coefficients for Interferometric Phase Linking
Magnus Heimpel *, Irena Hajnsek, Othmar Frey · 2025
Read Quality Coefficients for Interferometric Phase Linking by Magnus Heimpel *, Irena Hajnsek, Othmar Frey (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-24512.txt.
Introduction
Magnus Heimpel, Irena Hajnsek, and Othmar Frey, three authors, study radar phase data. InSAR uses radar images over time to track ground motion. Phase linking merges a pixel's phase history into one clean record. Scattered pixels give messy phase data that may not support the merge. Users need a score that says how much to trust each pixel.
The team propose three quality scores inside one shared framework. One score grades fit, from best case down to a noise floor. One score reads closure phase from the data before any merge. One score compares the merge with its best rival option. All scores run from zero to one, with one most trusted. Tests on radar data over Visp in Switzerland looked plausible.
Catalog record
| CREATOR | Magnus Heimpel *, Irena Hajnsek, Othmar Frey |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 110 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-24512
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Quality Coefficients for Interferometric Phase Linking Magnus Heimpel *, Irena Hajnsek, Othmar Frey Abstract In multi-temporal InSAR, phase linking refers to the estimation of a single-reference interferometric phase history from the information contained in the coherence matrix of a distributed scatterer. Since the phase information in the coherence matrix is typically inconsistent, the extent to which the estimated phase history captures it must be assessed to exclude unreliable pixels from further processing. We introduce three quality criteria in the form of coefficients, for threshold-based pixel selection: a coefficient based on closure phase that quantifies the internal consistency of the phase information in the coherence matrix; a goodness-of-fit coefficient that quantifies how well a resulting phase history estimate approximates the phase information according to the characteristic optimization model of a given phase linking method; and an ambiguity coefficient that compares the goodness of fit of the original estimate with that of an orthogonal alternative. We formulate the phase linking methods and these criteria within a unified mathematical framework and discuss computational and algorithmic aspects. Unlike existing goodness-of-fit indicators, the proposed coefficients are normalized to the unit interval with explicit noise-floor correction, improving interpretability across stacks of different size. Experiments on TerraSAR-X data over Visp, Switzerland, indicate that the closure phase coefficient effectively pre-screens stable areas, the goodness-of-fit coefficient aligns with and systematically generalizes established quality indicators, and the ambiguity coefficient flags solutions that fit well but are unstable. Together, the coefficients enable systematic pixel selection and quality control in the interferometric processing of distributed scatterers. Keywords: InSAR; Phase Linking; Distributed Scatterers; Interferometric Phase Quality 1 Introduction Interferometric Synthetic Aperture Radar (InSAR) allows the retrieval of surface motion based on measured phase differences between repeated SAR acquisitions. Its relevance lies in its unique ability to provide spatially dense measurements over large areas, with the potential to achieve up to millimeter precision in the line-of-sight direction under favorable conditions. The precision of the phase measurements depends strongly on the interferometric coherence between acquisitions. Decorrelation of the radar targets introduces uncertainty in the interferometric phase, which propagates to the estimated surface deformation. Point-like scatterers with high phase stability that dominate the backscatter within a resolution cell, also known as persistent scatterers (PS), serve as suitable pixels to retrieve surface motion [ 1 , 2 , 3 , 4 ] . Tomographic approaches extend the retrieval to cases in which more than one point-like target is present in a range-azimuth resolution cell [ 5 , 6 , 7 ] . However, in regions with vegetation, especially at shorter wavelengths such as X-band, such scatterers are typically rare. In these cases, statistical analysis of distributed scatterers (DS) [ 8 ] , which represent the combined return of multiple similar scatterers that are much smaller than a resolution cell, may be necessary to achieve sufficient coverage across the scene. DS targets tend to be more prone to decorrelation, and their interferometric phases are generally inconsistent across different acquisition pairs. Therefore, strategies are needed to extract usable phase information from them. A common approach is to use phase linking methods, which estimate a single-reference interferometric phase history from the phases of all possible acquisition pairs. Over the past two decades, a variety of such techniques have been developed. In essence, they define an objective function that quantifies the loss of information between the set of all phases of interferometric pairs and a candidate phase history and solve the resulting optimization problem. Among the most established methods are the Small Baseline Subset (SBAS) method [ 9 ] ; phase triangulation (PT) approaches such as those proposed by Monti-Guarnieri and Tebaldini [ 10 ] , Ferretti et al. [ 11 ] , and Cao et al. [ 12 ] ; and eigendecomposition-based methods such as EMI [ 13 ] and CAESAR [ 14 ] . Sequential estimators [ 15 , 16 ] extend these approaches to large-scale datasets by processing subsets of the data stack, which can improve scalability and allow efficient incorporation of new acquisitions. These methods face a common challenge: when applied with relaxed constraints, the resulting phase estimates may appear smooth yet be physically meaningless in strongly decorrelated areas. Consequently, robust criteria are needed to evaluate the reliability of phase linking outputs and to determine whether a given DS pixel should be included in subsequent processing. Previous works have introduced different quality indicators to address this issue. Ferretti et al. [ 11 ] proposed a phase residual metric for filtering unreliable points in the SqueeSAR framework. Samiei-Esfahany et al. [ 17 ] incorporated the same metric into their integer least squares estimator and Even and Schulz [ 8 ] discussed refining it with coherence-based weighting. Ansari et al. [ 13 ] derived a different indicator based on the eigenstructure of the coherence matrix. However, these criteria are typically tied to specific methods and no unifying framework has been proposed that can generalize across the different classes of phase linking algorithms. Moreover, the scale of these indicators is not normalized and typically depends on the number of acquisitions, which complicates thresholding and comparison across datasets. In this work, we develop such a framework that encompasses the most commonly used phase linking methods. Based on this formulation, we define several scalar quality coefficients that provide interpretable measures of reliability for DS phase estimates. Our specific contributions are: 1. In Section 2 , we extend the framework of Cao et al. [ 12 ] and show that a broad range of phase linking methods can be understood as approximations of the coherence matrix and characterized by a distance function. This unified perspective clarifies how existing methods relate and enables a consistent categorization. Closely related, the COFI-PL framework [ 18 ] introduced by Vu et al. interpreted phase linking as projecting an input covariance estimate onto the set of phase-consistent matrices. Our formulation overlaps with this perspective while encompassing a wider range of methods and offering a different categorization. 2. In Section 3 , we introduce a set of normalized scalar coefficients to evaluate phase linking results: • A closure phase coefficient quantifying the internal consistency of the interferometric phases. This coefficient is computed directly from the coherence matrix and independent of phase linking. • A method-specific goodness-of-fit coefficient that normalizes the optimization objective that characterizes each phase linking method, using theoretical and empirical bounds. It offers improved interpretability compared to existing goodness-of-fit indicators due to normalization and noise-floor correction. • A method-specific ambiguity coefficient representing the condition of the optimization problem. It compares the goodness-of-fit coefficients of the original solution and a secondary solution which is defined as the best solution among those that are orthogonal to the original. 3. In Section 4 , the proposed coefficients are evaluated on a TerraSAR-X dataset acquired over the city of Visp in the Swiss Alps, including both urban and vegetated areas. The dataset exhibits typical characteristics of high-resolution InSAR time series in challenging terrain, such as temporal decorrelation and strong spatial variation in the atmospheric phase component. It serves as a representative test case for validating the plausibility and interpretability of the proposed coefficients. 4. In Appendix C , we discuss numerical algorithms for phase triangulation as well as strategies for generating