What Are Infrared Image Sensors?

Infrared image sensors are detector arrays that convert two-dimensional distributions of infrared radiation into electronic images, enabling visualization of thermal emission patterns and near-infrared scene content that is invisible to the human eye. Unlike conventional silicon-based imaging sensors, which are sensitive to visible and near-infrared light below approximately 1.1 micrometers, infrared image sensors extend detection into the mid-wave (MWIR, 3–5 µm) and long-wave (LWIR, 8–14 µm) spectral bands where objects near room temperature emit their peak thermal radiation. The defining component of a modern infrared image sensor is the focal plane array (FPA), an integrated assembly of thousands to millions of individual detector elements arranged in a grid pattern, each producing a signal proportional to the infrared flux it receives.

Infrared imaging grew from single-element scanning systems used in the 1950s and 1960s toward staring arrays capable of producing a full image without mechanical scanning. NASA Technical Reports Server archives document early work on bolometer and photosensitive detector systems that laid the groundwork for modern FPA designs. The shift to two-dimensional staring FPAs drove a major expansion in infrared camera capabilities and, with the maturation of uncooled microbolometer technology in the 1990s, enabled compact and affordable thermal cameras suitable for commercial and consumer use.

Focal Plane Array Architecture

A focal plane array integrates an infrared-sensitive detector layer with a silicon readout integrated circuit (ROIC). The ROIC performs charge integration, multiplexing, and noise filtering for each pixel, converting the detector output into a serial digital stream that feeds downstream image processing electronics. Detector elements are bonded to the ROIC using indium bump bonding, a process in which tiny indium solder bumps connect each detector pixel to its corresponding readout cell, forming what is known as a hybrid FPA. Monolithic FPAs, where detector and readout circuitry are fabricated on the same substrate, are used for some uncooled thermal detector designs and offer advantages in manufacturing cost. Pixel pitches have decreased steadily over time, from 50 micrometers in the 1990s to 12 micrometers or smaller in current high-density arrays, increasing spatial resolution without growing the chip footprint.

Cooled and Uncooled Sensor Technologies

Cooled infrared image sensors use photon-detecting semiconductor materials, primarily mercury cadmium telluride (HgCdTe) and indium antimonide (InSb), that must be operated at cryogenic temperatures, typically 77 K, to suppress thermally generated carriers that would otherwise overwhelm the photon-induced signal. These sensors achieve high sensitivity and fast response but require a cryogenic cooler, increasing system size, cost, and power consumption. Uncooled sensors, by contrast, use thermal transducers, most commonly microbolometers based on vanadium oxide (VOx) or amorphous silicon, that respond to the temperature rise induced by absorbed radiation. Uncooled FPAs operate at ambient temperature and are smaller and less expensive, making them the dominant choice for handheld thermal cameras, building inspection tools, and automotive night vision systems. A detailed comparison of cooled and uncooled technologies for thermography applications has been examined in research published in the Journal of Nondestructive Evaluation.

Readout and Image Processing

The ROIC output is a stream of raw detector values that must be corrected and processed before being displayed or interpreted. Non-uniformity correction (NUC) applies per-pixel gain and offset calibration coefficients to compensate for the inevitable variation in detector responsivity and zero-level signal across the array; without NUC, fixed-pattern noise would dominate the image. Bad pixel replacement algorithms substitute estimated values for pixels that are permanently faulty or that have drifted outside acceptable calibration bounds. Analog-to-digital conversion within the ROIC or in an immediately downstream circuit produces a digital image stream, typically at 14-bit resolution for scientific-grade sensors. Further processing steps such as histogram equalization, false-color mapping, and image fusion are applied in downstream electronics. The IEEE Transactions on Electron Devices regularly publishes research on FPA device physics, ROIC design, and advanced readout architectures.

Applications

Infrared image sensors have applications in a wide range of disciplines, including:

• Thermal inspection of electrical systems, buildings, and industrial equipment for predictive maintenance
• Automotive night vision systems for pedestrian detection and collision avoidance
• Medical thermography for fever screening and vascular imaging
• Remote sensing satellites for land surface temperature mapping and wildfire monitoring
• Defense and security systems for surveillance, target acquisition, and search and rescue