A multi-snapshot inversion technique for interferometric Earth observation in the L-band

A computationally tractable multi-snapshot inversion technique for fusing visibilities in the interferometric domain is proposed to denoise and unfold interferometric images, increasing radiometric sensitivity and expanding swath width. Aliasing artifacts corrupt much of the field of view of images recovered from a set of visibilities computed over a single correlator integration time by antenna arrays with excessive element spacing. By fusing a sequence of visibilities computed over successive integration times before inversion, this proposed method can recover the emissivities of each pixel visible to the satellite in each snapshot.

• The computational tractability of interferometric fusion is made possible by a yaw-steering correction, like that of SMOS, which maintains constant pitch and induces invariance of acquisition geometry and surface parameters in geodetic coordinates with respect to the satellite's ground trace. Thus, sequences of visibilities can be obtained by convolving an image of the sparse brightness temperature parameters with a fixed observation function. This convolution can be diagonalized using the discrete Fourier transform, leading to a frequency-by-frequency inversion algorithm whose design matrices are asymptotically smaller than those used to perform snapshot-by-snapshot inversions.

• That the L-band directional emissivity patterns in ground-frame polarizations of patches of Earth's surface can be sparsely parameterized enables the multi-snapshot inversions to strongly denoise the recovered images, as the number of degrees of freedom of this parameterization does not grow with the number of snapshots.