CS Camera Technology

How a Compressive Sensing Camera Works

Compressive Sensing Camera

The operation of a single-pixel, single-diode, Compressive Sensing camera. InView has also designed multi-diode cameras.

New, Different, Better… at Prices that Change The Way People Work.

A Compressive Sensing camera replaces the spatial measurements normally made by a conventional of a Focal Plane Array (FPA) with a series of temporal measurements made by one or a small number of detector diodes. This significantly reduces the cost of infrared cameras.

The temporal measurements are fed into an algorithm that reconstructs the spatial information of the original image.

  1. In a Compressive Sensing camera, the incoming image information is focused onto a Spatial Light Modulator (SLM), more specifically a Texas Instruments Digital Micromirror Device (DMD). The DMD is positioned in the camera where a Focal Plane Array would normally be positioned in a conventional camera architecture.
  2. During the capturing of one image, a time series of unique SLM settings (DMD mirror patterns) are used sequentially (temporally). Each pattern selects one-half of the light of the original image. The exact mirror patterns used are prescribed by Compressive Sensing mathematical theory, and are orthogonal to the information in the image, allowing maximum information about the scene to be captured by the system
  3. The DMD micromirrors can be individually tilted left or right by 12 degrees. For each measurement, one-half of the mirrors are tilted in such a way that they reflect to a single photo-detector diode that portion of the incoming image which focused upon those mirrors. As a result, the total light energy of half of the image is concentrated by optics onto a single diode detector.
  4. The photons in half of the image are converted to an electrical signal by the diode, an amplifier and an analog-to-digital converter. A time series of measurements are created for the incoming image. Each individual measurement is the function of a specific mirror pattern.
  5. A reconstruction algorithm uses knowledge of the mirror patterns and the associated measured data to determine what image had to have been present at the input of the camera. The reconstruction algorithm is linear and has been mathematically proven to reproduce the original image.

Advanced Infrared Camera Capabilities

InView has developed both a single-diode camera, and a multi-diode camera . The multi-diode device uses a small number of didoes to significantly reduce the data collection time per frame.

Compressive sensing cameras support a dynamic tradeoff between resolution and frame rate. At low resolutions, frame rates can exceed that of conventional FPA –based infrared cameras.

InView has demonstrated an advanced “solar-exclusion” capability where the camera can dynamically identify a bright object present in the field of view, and block that bright, interfering light source from reaching the measurement diode.

Hyperspectral Imaging

InView also has experience with Hyperspectral Imagers (HSI) based upon Compressive Sensing, where each spectral window is captured by a separate single-diode camera. This architecture supports a staring HIS without use of a scanning mirror at the input. This approach also supports extremely fast spatial signature detection.

InView has demonstrated an adaptive HSI that can dynamically perform pixel aggregation for a target and eliminate background signal to improved overall signal-to-noise performance for a target of interest.

Intellectual Property

InView Compressive Sensing cameras are protected by 31 patents and/or patent applications, including US patents 8199244, 7271747 and 7511643.
Compressive Sensing Cameras

COMPRESSIVE SENSING CAMERAS

A Compressive Sensing camera replaces the spatial measurements normally made by a conventional of a Focal Plane Array (FPA) with a series of temporal measurements made by one or a small number of detector diodes. This significantly reduces the cost of infrared cameras.

The temporal measurements are fed into an algorithm that reconstructs the spatial information of the original image.