Low SNR Multiframe Registration for Cubesats

TL;DR

This paper introduces a fast, non-iterative registration algorithm for noisy multiframe images, optimized for low SNR conditions on resource-limited platforms like cubesats, enabling efficient on-board image fusion.

Contribution

The proposed algorithm is novel in its non-iterative design, requiring no hyperparameter tuning, and is optimized for embedded systems with low SNR image data.

Findings

Achieves accurate registration at very low SNR levels.

Requires only FFT, multiplication, and downsampling operations.

Optimal in maximum likelihood sense for certain noise models.

Abstract

We present a registration algorithm which jointly estimates motion and the ground truth image from a set of noisy frames under rigid, constant translation. The algorithm is non-iterative and needs no hyperparameter tuning. It requires a fixed number of FFT, multiplication, and downsampling operations for a given input size, enabling fast implementation on embedded platforms like cubesats where on-board image fusion can greatly save on limited downlink bandwidth. The algorithm is optimal in the maximum likelihood sense for additive white Gaussian noise and non-stationary Gaussian approximations of Poisson noise. Accurate registration is achieved for very low SNR, even when visible features are below the noise floor.