Ubicept's Physics-Based Imaging Redefines Machine Perception

BOSTON, MA – January 05, 2026 – Amid the dazzling displays of consumer electronics at CES 2026, a computer vision startup is tackling one of the most fundamental obstacles hindering the progress of artificial intelligence: the ability for machines to see. Ubicept, founded by experts from MIT and the University of Wisconsin-Madison, is showcasing a new imaging toolkit that promises to grant autonomous vehicles, robots, and augmented reality systems clear, reliable vision even in the most challenging conditions.

The company is providing live demonstrations of its technology, which uses a physics-based approach to dramatically reduce the noise and motion blur that plague conventional camera sensors. This enhancement applies not only to the ubiquitous CMOS sensors found in nearly every modern device but also to the next generation of single-photon avalanche diode (SPAD) sensors, which are poised to revolutionize machine perception.

In a significant demonstration of this new capability, Ubicept is partnering with Canon Americas Lab to showcase its processing software on Canon’s new high-frame-rate 1-megapixel SPAD sensor. The joint demonstration highlights how the technology enables the capture of crisp, clear images of fast-moving objects in difficult lighting, a critical need for industries from automotive to industrial automation.

The Physics of Sight: Beyond Conventional Imaging

For years, the solution to poor-quality images has been to apply more AI, using neural networks to denoise, deblur, and enhance video feeds. However, Ubicept argues this approach can be a double-edged sword for safety-critical systems. AI-based enhancement can sometimes generate details that look real but aren't physically accurate, creating potentially untrustworthy data for a self-driving car's perception system.

Ubicept’s technology takes a different path. Instead of post-processing a flawed image, it goes back to the fundamental physics of light. The Ubicept Toolkit is built around two core components: the Ubicept Photon Fusion (UPF) algorithm and the Flexible Light Acquisition and Representation Engine (FLARE) firmware. This system is designed to overcome what industry insiders call the "impossible triangle" of machine vision: the challenge of simultaneously optimizing for low light, fast motion, and high dynamic range (HDR).

UPF works by processing the raw data from the sensor with extreme temporal precision, effectively capturing and fusing individual photons over time to construct a clear image. This method preserves critical visual features like corners and edges, which are vital for AI perception tasks like simultaneous localization and mapping (SLAM) but are often smoothed over or lost by traditional denoising techniques. The result is a clean, physically grounded video stream that provides a more reliable input for AI decision-making.

While optimized for the nanosecond precision of advanced SPAD sensors, the company claims its Toolkit can also provide immediate image quality improvements for the billions of CMOS cameras already in operation, offering customers a clear upgrade path for existing hardware.

Powering the Next Generation of Sensors

Ubicept's collaboration with Canon puts a spotlight on the future of imaging hardware: SPAD sensors. Unlike traditional CMOS sensors that collect light over a set exposure time, SPADs are so sensitive they can detect and time the arrival of individual light particles, or photons. This provides an unprecedented level of detail about a scene, especially regarding motion and depth.

However, this sensitivity comes with a major challenge. A 1-megapixel SPAD sensor can generate over 100 gigabytes of data per second—a deluge of information that is impossible for most systems to process into a usable image in real-time. This data bottleneck has been a major barrier to the widespread adoption of SPAD technology.

This is where Ubicept's FLARE firmware becomes critical. It employs advanced, application-specific encoding schemes to manage the massive data stream, intelligently compressing the information without losing the essential details needed for accurate image reconstruction. By solving this bandwidth problem, Ubicept’s software effectively unlocks the full potential of SPAD hardware, making it practical for real-world applications.

The market is already shifting in this direction. While the CMOS sensor market continues to grow, analysts project the SPAD market is expanding nearly three times faster, signaling a major technological transition over the next decade. By positioning its software as the essential bridge to this next-generation hardware, Ubicept is placing itself at the center of this evolution.

Redefining Safety and Capability in Autonomous Systems

The most profound impact of this technology may be in the realm of autonomous systems, where perception failures can have catastrophic consequences. For self-driving cars, challenging lighting is a notorious weak point. The glare of oncoming headlights, the sudden darkness of a tunnel, or the low visibility of a rainy night can blind conventional sensors, leading to critical errors.

By delivering clear, blur-free imagery in high-contrast and low-light scenarios, Ubicept's technology directly addresses these failure points. This promises to make vision-based safety systems more reliable, accelerating the development and deployment of safer autonomous vehicles. The company is already working with automotive industry partners, including Tier 1 supplier Continental, to integrate these capabilities into future vehicles.

The benefits extend far beyond public roads. In industrial robotics, the technology enables robots to operate with greater precision and speed in dimly lit factory floors or warehouses. For high-speed manufacturing, it allows for more accurate defect detection on production lines without needing to slow down processes. In the world of AR/VR, it allows for more precise environmental tracking and 3D object modeling without the need for highly controlled studio lighting, paving the way for more seamless and immersive experiences on consumer devices.

Building an Ecosystem for Advanced Perception

Ubicept’s strategy is not to build its own cameras, but to create a broad ecosystem by partnering with both hardware manufacturers and the companies deploying autonomous systems. The collaboration with Canon is a prime example, as is a previously announced partnership with Pi Imaging Technology, a ZEISS company and leader in SPAD technology.

“We strive to work with camera manufacturers and robotics developers to achieve the next generation of smart perception,” said Sebastian Bauer, co-founder and CEO of Ubicept, in the company’s announcement. “With our recently released Toolkit for CMOS cameras, customers have more options than ever to see how Ubicept’s technology can supercharge their perception systems in robotics, automotive, and industrial sensing.”

This collaborative approach acknowledges that advancing machine perception requires synergy between hardware, software, and application-specific integration. The company is also working to ensure its software can run efficiently on widely used edge computing platforms from chipmakers like Qualcomm and NXP, further lowering the barrier to adoption. By creating a toolkit that enhances today's sensors while paving the way for tomorrow's, Ubicept is positioning itself as a key enabler for the entire industry. The demonstrations at CES 2026 are not just a product showcase; they are a glimpse into a future where machines can finally see the world as clearly as it needs to be seen.