Abstract

Our visual systems constantly gather and process data from the complex four-dimensional world we are immersed in. Streams of light, after intricate interactions within the environment, are reorganised as sequences of images by optical systems such as our eyes or cameras. We subconsciously interpret the pictures as two-dimensional snapshots of our world—a slice in time and a projection down from three-dimensional space. I will present my recent work on many aspects related to the synthesis of photo-realistic images given virtual descriptions of the scene as well as camera. I will argue that the notion of photo-realism is coupled with the ever-increasing sophistication of camera technology.

In this context, I will discuss empirical tools to analyse three problems of practical photo-realistic depiction: simulation, abstraction and representation. First, I will talk about a frequency domain analysis of light fields and demonstrate its utility in efficient Monte Carlo integration. Next, I will introduce empirical mode decomposition as a tool for detail abstraction. Finally, I will briefly present work in progress, on using multidimensional scaling for alternate image representations.

Bio:

Kartic Subr is a Newton fellow and honorary research associate at University College London. Previously, Kartic was a post-doctoral researcher and Marie-Curie fellow at INRIA -Grenoble, France. He received the doctoral degree for his work on sampling strategies for Monte Carlo integration, at the University of California, Irvine under the late Prof. James Arvo who was a pioneer of light transport methods in computer graphics.