ABSTRACT
Terahertz imaging with compressive sensing
by
Wai Lam Chan
Most existing terahertz imaging systems are generally limited by slow image ac-
quisition due to mechanical raster scanning. Other systems using focal plane detector
arrays can acquire images in real time, but are either too costly or limited by low
sensitivity in the terahertz frequency range.
To design faster and more cost-effective terahertz imaging systems, the first part
of this thesis proposes two new terahertz imaging schemes based on compressive sens-
ing (CS). Both schemes can acquire amplitude and phase-contrast images efficiently
with a single-pixel detector, thanks to the powerful CS algorithms which enable the
reconstruction of N-by-N pixel images with much fewer than N2 measurements. The
first CS Fourier imaging approach successfully reconstructs a 64 × 64 image of an ob-
ject with pixel size 1.4 mm using a randomly chosen subset of the 4096 pixels which
defines the image in the Fourier plane. Only about 12% of the pixels are required for
reassembling the image of a selected object, equivalent to a 2/3 reduction in acqui-
sition time. The second approach is single-pixel CS imaging, which uses a series of
random masks for acquisition. Besides speeding up acquisition with a reduced num-
ber of measurements, the single-pixel system can further cut down acquisition time
by electrical or optical spatial modulation of random patterns.
In order to switch between random patterns at high speed in the single-pixel