Compressive Sensing

Compressive Sensing is an innovative technology that reduces the cost of Shortwave Infrared (SWIR) cameras by replacing expensive InGaAs focal plane arrays with low cost components and signal processing algorithms.

InView’s Compressive Sensing technology builds upon $10M of research performed at Rice University, protected US patents: 8,199,244, 7,271,747 and 7,511,643.  InView’s patented CS technology is also covered by US patents: 9,277,139 ; 9,160,914 ; 9,081,731 ; 8,970,740 ; 8,922,688 ; 8,885,073 ; 8,860,835 ; 8,760,542 ; 8,717,551 ; 8,717,492 ; 8,717,484; 8,717,463; 8,717,466 ; 8,634,009; 8,570,406; 8,570,405

Compressive Sensing (CS) is an innovative data sampling strategy that exploits the fact that images are often sparse in some transform basis to perform imaging and machine vision processing tasks from far fewer measurements than required by traditional Nyquist sampling rules.  

Inside the InView210 high-resolution scientific SWIR camera

The InView210 high-resolution scientific SWIR camera

Compressive sensing gives the InView210™ platform a very different architecture than common digital cameras.  It is a unique combination of optical components, electronics and algorithms.  In compressive sensing, a scene is observed using a measurement matrix implemented row by row as patterns displayed on a high-speed digital micromirror array. Each pattern represents a basis function from a set that optically projects the scene onto a mathematical representation that significantly reduces its dimensionality without losing information. The scene is modulated then focused to a single detector, as shown below.   Using proprietary processing techniques, the low dimensional data vector constructs high-resolution images from measurements numbering only a fraction of the number of pixels that are constructed.  InView is now demonstrating that new compressed-domain machine vision algorithms can automatically detect and classify targets from even fewer measurements. InView proposes to apply these new processing techniques to efficiently find information in large digital data sets for high-speed pattern recognition and diagnostics.


For more information see our references:

  1. A. Herman, et al., “Recent results in single-pixel compressive imaging using selective measurement strategies,” Proc. SPIE v. 9484, 9494-09 (2015).
  2. A. Herman, et al., “A higher-speed compressive sensing camera through multi-diode design“, Proc. SPIE v 8717, 871706 (2013).
  3. Herman, “Compressive Sensing with Partial-Complete, Multiscale Hadamard Waveforms,” in Imaging and Applied Optics, (2013), paper CM4C.3.
  4. Lenore McMackin, “Design of a multi-spectral imager built using the compressive sensing single-pixel camera architecture,” of SPIE Vol. 9761, 97610G (2016).
  5. Herman, “Computational imaging with signature row-block Kronecker products,” IEEE Signal Processing Mag. (to be published).