Today, smartphones are good enough to shoot professional stock photography. But the images produced by the first phone-mounted cameras were once laughingstock photography. Moore's Law, multiple lenses, and AI-driven algorithms have since combined to deliver incredible images, with particularly profound advances over the past few years in dynamic range and low-light capture that have amounted to an open war on the DSLR. Despite the impressive capabilities of today's smartphones, though, more advanced imaging capabilities lie ahead both for general photography and computer vision.
One example is technology from SeeDevice called PAT-PD (Photon Assisted Tunneling-Photo Detector), a way of practically eliminating "photon leakage" from a sensor that the company says results in a mix of visible and near-infrared photo capabilities. The stated benefits include moving beyond today's sensors in such areas as reducing motion distortion, low-light capture, and dynamic range. The company has posted several comparison images on its website, including one where it outperforms a DSLR in an extreme low-light still image.
SeeDevice says its technology is so powerful that it can capture images of blood vessels under the skin, opening the way for biometric and biomedical applications (and cool horror movies). "Depending on the wavelength we tune to," says VP of Marketing Mark Shedd, "we can detect anything from glucose levels, fat/cholesterol levels, heart rate, and even alcohol level." The company also offers that it can capture far greater clarity at a greater distance, which could open it up to autonomous automotive applications. Indeed, vehicles might see such technology deployed both internally and externally with computers monitoring the road ahead for potential dangers and internally to monitor the well-being and mood of drivers or passengers.
SeeDevice's approach isn't the only option for extreme low-light imaging. Sionyx, for example, has expanded its Aurora low-light camcorder into a family of products aimed at applications that include search-and-rescue, surveillance, and hunting. The company's technology uses a technology called "black silicon," which SeeDevice says it can outperform. In automotive applications, optical cameras also must vie with lidar offerings from companies such as Luminar, Velodyne, and Bosch. But, while the company is exploring these kinds of applications, it maintains that its sensors are also ideal for smartphones, noting that it bucks the trend of requiring larger photodetectors to maximize light capture. In fact, it says that its performance improves as its photodetectors become smaller.
Other imaging innovations can help identify objects without necessarily improving image quality. For example, earlier this year, I wrote about Trinamix, a BASF company that had developed an algorithm based on backscatter reflections to reliably discern among a wide range of materials even if they have identical colors and shapes. The company was making a bid to add security to facial login systems by ensuring that identified faces had live skin rather than being a convincing mask. In contrast, the SeeDevice sensor isn't commercially available yet. The company recently announced a licensing agreement with MegaChips Corporation, enabling the fabless silicon developer to integrate SeeDevice's technology into its products. SeeDevice expects its sensors to be available in commercial products next year.