Abstract
<title>Abstract</title> <p>Astronomical image analysis commonly uses spectral, statistical, and model-based tools to characterize faint structures, suppress noise, and compare image fields. These methods are powerful, but Fourier amplitude statistics alone do not preserve the phase-dependent spatial organization that determines connectivity, anisotropy, and persistence. This paper introduces an informational coherence-geometry pipeline for astronomical images, inspired by Viscous Time Theory but implemented entirely with reproducible public operations: grayscale normalization, structure-tensor coherence mapping, artifact-aware masking, Fourier-spectrum-preserving phase-randomized surrogates, and connected-component analysis. The method is tested on two public JWST NIRCam image products: the COSMOS Legacy Field and Cassiopeia A. In both targets, real images produce fewer but substantially larger connected high-coherence regions than phase-randomized surrogates with the same Fourier amplitude spectrum. In COSMOS, the largest real cluster contains 1,358 pixels, compared with a surrogate ensemble maximum of 223 pixels; in Cassiopeia A, the corresponding values are 780 and 285 pixels. In both cases, the empirical ensemble significance is p=1/(50+1)≈0.0196. These results do not identify new astrophysical objects. They show that coherence topology is a useful phase-sensitive descriptor of spatial organization beyond spectral equivalence.</p>