About:

Digital images permeate many aspects of our lives, from our social interactions and entertainment, to our food, healthcare, and security systems. Images are also central to scientific discovery, from the exploration of the cosmos to the inner workings of our minds and our biology. These ubiquitous images are far more than snapshots captured by pocket cameras; they are carefully crafted creations born from a marriage between sophisticated and specialised devices (e.g., satellite-borne telescopes, medical imaging scanners, defence radars), and cutting-edge mathematical and computational techniques. Behind every image lies decades of advanced mathematical research in the form of computer algorithms that analyse raw sensor data to produce high-quality images and extract information from them. At the heart of these algorithms sit mathematical models that describe what we consider to be a high-quality image and that allow the algorithms to separate signal from noise. But how to describe what images look like without letting our own preconceived ideas, opinions, and views negatively bias results? Can we rely on artificial intelligence to derive accurate image models from image data, without compromising on the fairness, accountability, and transparency of our imaging technologies? In this talk, Marcelo Pereyra looks at some of the key mathematical ideas that have enabled modern imaging technology, focusing on the complex and multifaceted role that bias plays in this fascinating mathematical and engineering endeavour.