Statistics in Imaging Science
Forty-thousand years ago, human ancestors documented their history on stone with hieroglyphics. Over five-hundred years ago, Leonardo da Vinci painted the Mona Lisa. Today, discovery and innovation depend on the production of digital images while social media allows people around the globe to share online more than three billion digital photos per day.
The use of images to communicate cultures and express creativity has evolved over thousands of years to envelop scientific research. Telescopes capture images of galaxies millions of light-years away, mesoscale microscopes can show high-resolution images at the subcellular level, and satellite images enable us to monitor the atmosphere and predict extreme weather.
Imaging science is about the development of theory, methods, and systems for creating better images, communicating, and storing them efficiently, finding patterns, and interpreting image content automatically. These research problems have been pursued for decades in a variety of fields, for instance, image processing, computer vision, and machine learning. Imaging science presents opportunities for data scientists to collaborate across disciplines and impact drastically different academic fields - biology, medicine, meteorology, material science, and art history, to name a few.
In our department, faculty members work on diverse topics in imaging science. Dr. Jia Li has worked on image processing (e.g., segmentation), image retrieval, image annotation, aesthetics assessment, and computational approaches to art history studies. Dr. Nicole Lazar has worked on the statistical analysis of brain images, to deepen our understanding of mental illness as well as healthy development. Challenges in imaging science have motivated them to develop new statistical techniques: methods for multiple testing, large-scale spatiotemporal models, probabilistic graph models, and new algorithms for supervised and unsupervised learning which are applicable to more general data.