mm/Sub-mm Wave Compressive Sensing Imaging

Imaging in the millimeter-wave to terahertz (mmw-THz, 30GHz-3THz) region of the electromagnetic spectrum provides a new dimension to the investigation and perception of the world around us. Millimeter wave imaging has been extensively used for threat detection at airports and for scientific discovery in radio astronomy to identify particles in gas clouds thousands of light years away. Terahertz imaging has shown a new means to investigate biological tissues, in particular in tumor screening. While the scientific and societal benefits are enormous, the field of mmw-THz imaging is still in its infancy. A particularly acute challenge is that it is impractical to integrate a large number of mmw-THz sensors on a circuit board, because the size of each 'pixel' must be at least proportional to the wavelength. Unfortunately, using a single sensor or small number of sensors renders current mmw-THz imaging systems far too slow. In this work we will investigate a new method for mmw-THz imaging that uses compressive sensing and a digital mmw-THz configurable lens to provide up to a 32x improvement in resolution or speed.

Compressive sensing allows a reduction in the information needed to reconstruct an image, thus enabling optical imagers that require 70-80% fewer measurements. Key to this reduction is the use of pseudorandom spatial modulators, or masks. In the mmw-THz frequencies, however, such spatial modulators do not exist, due to the size of each modulator 'pixel' needing to be at least proportional to the wavelength. Consequently, mmw-THz has not been able to leverage these new approaches. In this project, we will investigate a transformative solution to mmw-THz imaging by combining several research areas. First, photon induced dynamic spatial modulators will be used for mmw-THz frequencies. Second, advanced compressive sensing algorithms will reconstruct high fidelity images from a reduced number of measurements. Third, multi-sensor arrays will accelerate the image acquisition, and signal processing algorithms will account for the shared structure of the image being acquired by all the sensors. The resulting system has the potential to provide speeds of 50-100 frames per second and greatly increased signal-to-noise ratio.