The arrival of LANL's Roadrunner machine, the world's first petascale computer capable of over 1015 floating point operations per second, matches some estimates of the computing power of visual processing regions within the human brain (visual cortex). The goal of research is now to understand what and how the visual cortex computes. Neuroscience describes the visual cortex in terms of many specialized functional regions, each consisting of sheets of cortical columns, each containing layered assemblies of tens of thousands of neurons. Each neuron is connected to thousands of other neurons in feed-forward, lateral, and feed-back circuits. We describe new work at LANL, funded by DOE and NSF, to build and demonstrate functional models of visual cortex that integrate the latest theories of neuron and column cortical processing, and apply these models to a range of computer vision applications.