Single-pixel imaging via compressed sensing can reconstruct high-quality images from a few linear random measurements of an object/scene known a priori to be sparse or compressive, by using a point/bucket detector without spatial resolution. Nevertheless, it still faces a harsh trade-off among the acquisition time, the spatial resolution and the signal-to-noise ratio. Here we present a new compressive imaging approach with use of a strategy called cake-cutting which optimally reorders the deterministic Hadamard basis. By this means, the number of measurements can be dramatically reduced by more than two orders of magnitude. Furthermore, by exploiting the structured characteristic of the Hadamard matrix, we can accelerate the computational process and simultaneously reduce the memory consumption of storing the matrix. The proposed method is capable of recovering an image of the object, of pixel size $1024\times1024$, with a sampling ratio of even 0.2%, thereby realizing super sub-Nyquist sampling and significantly reducing the acquisition time. Moreover, through the differential modulation/measurements, we demonstrate this method with a single-photon single-pixel camera under low light condition and retrieve clear images through partially obscuring scenes. This described practical method complements the single-pixel imaging approaches and can be applied to a variety of fields, such as video, night vision goggles and automatic drive.
History
Disclaimer
This arXiv metadata record was not reviewed or approved by, nor does it necessarily express or reflect the policies or opinions of, arXiv.