On Model Space Sampling in ADMIRE for Image Quality and Computational Efficiency

Image quality and computational efficiency of the aperture domain model image reconstruction (ADMIRE) method is dependent on the specific model used. This makes it important for users to understand how various model design considerations could have significant impacts on image processing. In this work, we consider how undersampling, randomization of physical locations, and dimensionality reduction with independent component analysis (ICA) can be used to improve runtime. Specifically, we observed that undersampling resulted in a trade-off between contrast and efficiency, random sampling led to improved CNR, and ICA generally showed improved contrast and computation time.