Apply for the development board now

Visionary.ai makes low-light video clear as day

Getting clear images in low-light conditions is a challenge. Videos taken at nighttime are often grainy and suffer from poor contrast, making it difficult to identify people or objects in frame. More-expensive sensors are an option, or you could just use the camera you have and Visionary.ai’s denoiser software…

When the ipXchange team visited The Valley for Sensors Converge 2024, we also chatted to David from Visionary.ai, a startup that is revolutionising AI-enhanced imaging in low-light environments. Visionary.ai’s denoiser software technology can be easily integrated into embedded systems for dramatically cleaner nighttime video recording. The cost, processing requirements, and power consumption are also minimal, which means all this capability can be done at the edge for real-time user experiences, like filming with a smart phone.

As David explains, Visionary.ai’s denoiser works by enhancing the raw sensor (camera) output and removing noise before this data enters the typical image processing pipeline of an embedded system. This means that Visionary.ai’s solution also allows you to use a cheaper camera and still get amazing image clarity. Alternatively, you can take a military-grade sensor and turn it up to 11.

Why remove image noise?

Image sensing when there is minimal light is a tricky business. All image sensors have a noise element to the data, but since the signal is determined by incoming light, dark environments suffer from having a very low signal-to-noise ratio – i.e. the noise level is comparable to the small amount of light coming in. This ultimately results in a poor-quality image.

By cleaning up this noise at the earliest point in the signal chain, Visionary.ai’s denoiser greatly enhances the potential of the final image after post-processing. The result is a nighttime video or photo that does not suffer from the randomised grain that spoils the image. Instead, an image much closer in clarity to one taken during the daytime is achieved.

In applications such as industrial monitoring and security, this level of high-quality, low-light imaging could be a gamechanger for keeping things safe as well as better monitored. Though the value of Visionary.ai’s solution for consumer end users and professional recording/photography is a huge plus, this technology has vast potential across industrial and security applications.

Running Visionary.ai’s denoiser

The ideal environment for integrating Visionary.ai’s software is a system that includes some form of neural networking capability – this is an AI solution. With that said, this can be a small neural processing unit within a larger chip or a smaller microprocessor or microcontroller, or it can be on dedicated AI hardware.

David highlights that many innovations that ipXchange has seen at Embedded World – from companies like Renesas and Alif – should be very capable of running Visionary.ai’s software. These microcontrollers are far lower cost and have far lower power consumption than typical processors, so you need not worry that Visionary.ai is just for the most heavy-weight AI architectures.

Visionary.ai’s denoiser in action

We are then treated to a demo of Visionary.ai’s technology in action. A portable darkroom environment – with moving objects inside – shows the software running on a Snapdragon SoC, with a ‘control’ phone serving as the same sensor setup without this denoiser installed. For a 50-megapixels sensor, the image latency is only half a frame.

This demonstration illustrates just how fast Visionary.ai’s denoiser works even on a relatively large image sensor. It also proves its use for real-time environments and user experiences, with vastly clearer images than when using a standard setup. In this instance, the non-enhanced system can barely register movement of the objects inside the box.

David also takes another phone with a much more expensive camera and shows how Visionary.ai’s solution still outperforms this image sensor even though the hardware is a fraction of the cost.

If you want to learn more about Visionary.ai and how to evaluate this technology, get connected by following the link below to our user-defined board page. There, you can apply for a consultation – specifying your project requirements and Visionary.ai – and ipXchange will put you in touch. We look forward to seeing where you’ll integrate this exciting AI innovation.

Keep designing!

User-Defined Evaluation Board Or Sample

Need crisp images for video data analysis in low-light conditions?

Apply for the development board now
Get industry related news

Sign up for our newsletter and get news about the latest development boards direct to your inbox.

We care about the protection of your data. Read our Privacy Policy.