Denoising Generalized Expectation-Consistent Approximation for MR Image Recovery
Authors:
Saurav K. Shastri,
Rizwan Ahmad,
Christopher A. Metzler,
Philip Schniter
Abstract:
To solve inverse problems, plug-and-play (PnP) methods replace the proximal step in a convex optimization algorithm with a call to an application-specific denoiser, often implemented using a deep neural network (DNN). Although such methods yield accurate solutions, they can be improved. For example, denoisers are usually designed/trained to remove white Gaussian noise, but the denoiser input error…
▽ More
To solve inverse problems, plug-and-play (PnP) methods replace the proximal step in a convex optimization algorithm with a call to an application-specific denoiser, often implemented using a deep neural network (DNN). Although such methods yield accurate solutions, they can be improved. For example, denoisers are usually designed/trained to remove white Gaussian noise, but the denoiser input error in PnP algorithms is usually far from white or Gaussian. Approximate message passing (AMP) methods provide white and Gaussian denoiser input error, but only when the forward operator is sufficiently random. In this work, for Fourier-based forward operators, we propose a PnP algorithm based on generalized expectation-consistent (GEC) approximation -- a close cousin of AMP -- that offers predictable error statistics at each iteration, as well as a new DNN denoiser that leverages those statistics. We apply our approach to magnetic resonance (MR) image recovery and demonstrate its advantages over existing PnP and AMP methods.
△ Less
Submitted 7 September, 2022; v1 submitted 8 June, 2022;
originally announced June 2022.
Generalized Design of Sampling Kernels for 2-D FRI Signals
Authors:
Saurav Kumaraswami Shastri,
Sunil Rudresh,
Chandra Sekhar Seelamantula
Abstract:
One of the interesting problems in the finite-rate-of-innovation signal sampling framework is the design of compactly supported sampling kernels. In this paper, we present a generic framework for designing sampling kernels in 2-D. We consider both separable and nonseparable kernels. The design is carried out in the frequency domain, where a set of alias cancellation conditions are imposed on the k…
▽ More
One of the interesting problems in the finite-rate-of-innovation signal sampling framework is the design of compactly supported sampling kernels. In this paper, we present a generic framework for designing sampling kernels in 2-D. We consider both separable and nonseparable kernels. The design is carried out in the frequency domain, where a set of alias cancellation conditions are imposed on the kernel's frequency response. The Paley-Wiener theorem for 2-D signals is invoked to arrive at admissible kernels with a compact support. As a specific case, we show that a certain separable extension of the 1-D design framework results in 2-D sum-of-modulated-spline (SMS) kernels. Similar to their 1-D counterparts, the 2-D SMS kernels have the attractive feature of reproducing a class of 2-D polynomial-modulated exponentials of a desired order. Also, the support of the kernels is independent of the order. The design framework is generic and also allows one to design nonseparable sampling kernels. To this end, we demonstrate the design of a nonseparable kernel and present simulation results.
△ Less
Submitted 19 February, 2019; v1 submitted 7 February, 2019;
originally announced February 2019.
