Imaging & Vision

Research Agenda: 

Imaging and Vision emerge from several critical areas including Medical Imaging (e.g., MRI, CT Scans), Exploratory Imaging (e.g., surveillance, space programs), Industrial Imaging (e.g., automated manufacturing monitoring), Artificial Vision (e.g., Google cars), and the multi-billion industries of animated movies, computer graphics, and video games. It also connects directly to the neural science on brains and human vision, as well as the computer science on A.I. As such, it is being actively pursued by engineers, medical practitioners, mathematicians, statisticians, and other scientists across multiple disciplines.

Our approach to Imaging and Vision has been driven by the “system” view, with main focuses on these areas.

• The graph, combinatorial, or relational structures for the system of  2D/3D  “objects” captured in a single or multiple images, including uncovering the hidden physical or cognitive organizations via optimal estimation.

• Biological or artificial “neural” network  systems that can reveal and emulate the efficiency (in speed), accuracy (in cognition), and robustness (in noisy or uncertain environments) of human vision decision making.

• Temporal image systems, and the associated optimal dynamic decision making in the framework of Bayesian Decision Theory and Hidden Markov Chains, with such applications as medical diagnostics and surgeries,  and unmanned aerial/ground vehicles.

• Image ensembles captured by dynamic robot or sensor networks, and their information fusion and diffusion, and their coupling with other type of sensor information such as infra-red, ranging,  or acoustic signals, for self-organized surveillance, navigation, or other coordinated tasks.

Such a systemic view also boosts our interest in multi-agent systems, including flocking systems.