The Limits of Conventional Imaging

At its core, machine vision depends on collecting photons reflected from a scene. For most of the history of industrial imaging, improving inspection performance has largely meant collecting more of them. Brighter illumination, larger optics, longer exposures, and more capable sensors have all contributed to better image quality and, ultimately, better inspection performance.

This strategy remains highly effective, but some inspection applications are beginning to encounter practical limits. High-speed inspection illustrates one example. Whether the motion originates from a production line, a robotic manipulator, a moving camera, or the part itself, reducing motion blur generally requires very short exposure times. The traditional solution is to compensate with additional illumination, often delivered through precisely synchronized strobe lighting. While effective, maintaining performance at increasing speeds requires more intense illumination and tighter synchronization between the camera, lighting, and motion of the target.

Speed is not the only factor that can limit image acquisition. Inspection systems are increasingly being deployed outside traditional machine vision cells, particularly in warehouse automation, robotic inspection, and large-scale manufacturing environments. In these applications, cameras may be required to observe targets from several meters away. As working distance increases, maintaining sufficient signal becomes more difficult, frequently requiring larger optics, more illumination, or both. In some environments, these solutions become expensive, physically impractical, or difficult to deploy at scale.

Some of the most difficult inspection tasks involve reflective materials. Polished metal surfaces, coated components, electronics, and packaging materials frequently contain both extremely bright and extremely dark features within the same scene. A defect may exist immediately adjacent to a specular reflection, making it difficult to capture all relevant scene detail in a single exposure. Increasing illumination can sometimes intensify reflections while doing little to reveal the underlying defect, while HDR techniques may introduce artifacts or struggle when scenes contain large dynamic range variations at production speeds.

These limitations become even more pronounced when inspection systems leave carefully controlled environments. Mobile robots and autonomous platforms often operate under changing lighting conditions and cannot always rely on precisely engineered illumination. Instead of controlling the scene, the imaging system must adapt to it.

For manufacturers, these imaging limitations manifest as measurement uncertainty. Blur, saturation, noise, and insufficient signal can increase the likelihood of missed defects, false rejects, and inconsistent inspection results, regardless of how sophisticated the downstream analysis may be.

Photon Counting Comes to Machine Vision

These requirements have motivated growing interest in alternative image sensor architectures, including photon-counting image sensors.

Virtually all machine vision cameras today rely on image sensors that accumulate electrical charge generated by incoming photons during an exposure period. At the end of the exposure, that accumulated charge is measured and converted into a digital value. This architecture forms the foundation of modern CMOS imaging and has delivered tremendous advances in performance over the past two decades.

Photon-counting sensors take a different approach. Rather than measuring accumulated charge, they detect individual photon arrivals and record those detections directly. One of the most promising implementations of this concept is the single-photon avalanche diode, or SPAD. When a photon reaches a SPAD pixel, it triggers an avalanche event that can be detected digitally, allowing the sensor to record the arrival of individual photons as they occur.

SPAD technology is not new. For years, SPAD sensors have been widely used in time-of-flight systems, LiDAR, and depth-sensing applications where precise photon timing is essential. Their ability to detect individual photons with exceptional temporal resolution has made them particularly valuable for ranging applications. What has changed in recent years is the emergence of higher-resolution SPAD arrays designed specifically for imaging.

Recent developments illustrate how quickly the technology is advancing. In 2025, Pi Imaging introduced the SPAD Alpha, a 1-megapixel SPAD camera capable of burst frame rates up to 57,000 frames per second. In the same year, Canon announced a 2.1-megapixel SPAD sensor architecture capable of approximately 156 dB of dynamic range, highlighting the potential of photon-counting approaches for scenes that push conventional imaging systems to their limits. These systems demonstrate that photon-counting imaging is rapidly evolving from a specialized sensing technology into a practical imaging platform.

The significance of these developments extends beyond sensitivity alone. Because photon detections are inherently digital, photon-counting sensors avoid the read-noise limitations associated with measuring extremely small analog signals. More importantly, they preserve the timing and distribution of photon arrivals that are largely unavailable in conventional image sensors. This capability enables imaging systems to retain measurements that are typically discarded during conventional image formation.

From Image Capture to Image Reconstruction

Traditional machine vision systems generally treat image acquisition and image analysis as separate processes. A camera captures an image, and subsequent algorithms operate on the resulting pixels. The image itself is assumed to be a direct measurement of the scene.

Photon-counting architectures create the possibility of a different approach. Because individual photon detections are preserved, measurements from many high-speed observations can be combined computationally before a final image is produced. Rather than treating image formation as a fixed process performed entirely by the sensor, systems are increasingly combining sensing and computation to reconstruct images from streams of photon detections.

Consider a high-speed inspection application in which object motion is known or can be estimated. A conventional imaging system must continuously balance motion blur against signal quality. Longer exposures collect more light but blur fine features. Shorter exposures freeze motion but reduce the available signal. Photon-counting architectures combine measurements from many high-speed observations while accounting for scene motion. At frame rates approaching tens of thousands of frames per second, dozens of observations can be accumulated within a fraction of a millisecond, allowing signal to be integrated without requiring a conventional long exposure.

Similar concepts can also be applied to robotic inspection systems, moving cameras, and other dynamic environments. In these scenarios, performance may depend not only on sensor capability and illumination, but also on the ability to combine measurements intelligently across time. As a result, image formation is now becoming more flexible than what is possible with conventional frame-based architectures.

Looking Ahead

The most difficult inspection problems are often the ones where critical signal is lost before an inspection algorithm ever evaluates an image. The importance of photon-counting sensors lies not simply in improved sensitivity, but in their ability to preserve and utilize measurements that conventional imaging systems may discard. As manufacturers continue pushing inspection systems toward higher speeds, larger workspaces, and more demanding operating environments, that capability may help reduce uncertainty at the earliest stage of the inspection process.

For quality and manufacturing leaders, the significance of photon-counting imaging extends beyond any single sensor specification. Emerging inspection requirements are increasingly being defined by limitations that cannot always be solved through additional illumination, larger optics, or incremental improvements in conventional image sensors. Organizations that understand these trends early will be better positioned to leverage next-generation imaging for a competitive edge.

Whether applied to high-speed inspection, challenging HDR scenes, or future computational imaging systems, photon-counting imaging is expanding the range of inspection problems that can be addressed with confidence.