Title: Bayesian image analysis in transformed spaces (BITS) and the BIFS/WIMP Python packages
Authors: John Kornak - University of California, San Francisco (United States) [presenting]
Karl Young - University of California San Francisco (United States)
Abstract: Bayesian image analysis can improve image quality by balancing a priori expectations of image characteristics with a model for the noise process. We will give a reformulation of the conventional image space Bayesian image analysis paradigm into Fourier and wavelet spaces. By specifying the Bayesian model in a transformed space, spatially correlated priors, that are relatively difficult to model and compute in conventional image space, can be efficiently modeled as a set of independent processes in an appropriately transformed space. The originally inter-correlated and high-dimensional problem in image space is thereby broken down into a series of (trivially parallelizable) independent one-dimensional problems. We will describe and show examples of the Bayesian image analysis in transformed space (BITS) modeling approach for both Fourier (BIFS) and wavelet (WIMP) space using both parametric and data-driven priors. In the process, we will showcase our Python package(s): BIFS/WIMP that can allow easy and fast implementation of BITS.