Recent developments in nano-technology and life sciences are creating structures on the scale of fractions of micrometers, down to the nanometer scale. Semiconductor manufacturing technology is successfully shrinking critical dimensions down to scales well below 100 nanometers. These trends in technologies require extremely high precision analysis tools for measurement, and visual feedback. We are moving into a world were structures ultimately need to be resolved with a resolution of single atoms, and molecules. Quantitative measurements are becoming essential for experimental progress, and manufacturing technology. However, even the most sophisticated imaging sensors, such as Scanning Probe Microscopes, or Scanning Beam Microscopes, are operating at the physical limits of resolution, and create data sets with extremely low spatial resolution, and low S/N. Nano-scale image data suffers from a number of problems: Noise and insufficient spatial sampling are the main sources of errors and can render even visual analysis by human observers impossible. Furthermore, structures, and dynamics of nano-scale objects exhibit a high degree of randomness imposed by the natural variability of atomic surfaces, and stochastic behavior. Quantitative analysis of 3D nano-structures, and their dynamics is a challenging task that can only be solved by an interdisciplinary approach, combining model-based knowledge from application fields, statistics, and state-of-the art image (sequence) analysis. This talk will introduce a research effort in Computational Nano-Vision that was recently founded as one of Intel’s advance research projects in Intel Research. We will outline ongoing research, and illustrate our approach with select examples of model-based data analysis techniques.
Copyright. All Rights Reserved.