Katie Bouman

Katherine L. Bouman is an American imaging scientist and computer vision researcher who led the development of the algorithmic pipeline that produced the first image of a black hole from Event Horizon Telescope data. Her CHIRP (Continuous High-resolution Image Reconstruction using Patch priors) algorithm uses sparse modeling and compressed sensing techniques to reconstruct images from sparse interferometric measurements — a problem that is mathematically ill-posed because the EHT's baselines sample only a small fraction of the spatial frequency domain.

Bouman's work exemplifies a critical but underappreciated aspect of modern observational science: the imaging pipeline is not a neutral post-processing step but an epistemic choice. Different algorithms encode different priors about what a reasonable astronomical source looks like, and these priors shape the resulting image in ways that are not always transparent. The EHT collaboration's practice of comparing multiple independent imaging pipelines — including those developed by Bouman's team and others — represents a nascent form of algorithmic triangulation that the field of computational imaging has not yet formalized.

Before her work on the EHT, Bouman contributed to the development of methods for seeing through scattering media, a problem with applications in medical imaging and terrestrial vision. Her broader research agenda treats imaging not as photography but as inference: given incomplete, noisy, and ambiguous measurements, what can be reliably concluded about the underlying scene? This framing connects her work to Bayesian inference, inverse problem theory, and the philosophy of scientific evidence.

The tendency to treat algorithmic reconstructions as transparent photographs — to forget that the image was selected from a space of possible images by a pipeline that makes assumptions — is not merely a public-relations problem. It is a methodological vulnerability. Bouman's insistence on cross-validating pipeline outputs is a first step toward making imaging science epistemically rigorous. The next step is to make those priors themselves objects of scientific debate, not hidden parameters.