How Smarter Cameras Could Help Change the World

Researcher Haijin Zeng has been working on a big question: how can we make cameras a lot smarter, especially in smartphones and scientific tools? His PhD research at Ghent University dives into improving a field called computational imaging – that’s a mix of camera hardware and smart computer tricks to make images clearer, more detailed, and even useful beyond what our own eyes can see.

“We don’t just want to take pretty pictures – we want to help computers understand the world through vision,” Zeng explains.

Seeing What the Eye Can't

Zeng worked especially on hyperspectral and multispectral cameras. These don’t just capture red, green, and blue like normal cameras, but hundreds of subtle colors (wavelengths). That helps detect things like early crop disease, air pollution, or even early signs of illness – things regular photos miss.

The challenge? These cameras collect a huge amount of data—far more than traditional systems can handle quickly and efficiently. To address this, Zeng developed an intelligent imaging pipeline: the camera samples sparsely to reduce raw data, and then smart algorithms reconstruct and “fill in” missing parts, even when inputs are noisy or incomplete. He combined mathematical techniques like tensor completion (a method for handling multi-dimensional data) with AI tools such as diffusion models (ChatGPT like generative models) to enable the camera to produce high-quality images rapidly without requiring labeled data.

“We designed methods that don’t need huge training datasets and that still give sharp, detailed results – even with noisy or incomplete input,” Zeng says.

From Drones to Smartphones

These breakthroughs aren’t just for labs. Zeng’s methods also improve small mobile cameras – like the one in your smartphone. He focused on Quad Bayer sensors, which are great in low light but usually give blurry results. His solution? A combined system that restores detail and removes noise at the same time – meaning clearer night photos on your phone, and better images from mini-cameras on drones or in medical devices.

A Boost for Smarter AI

Zeng’s research is also a step forward for Artificial General Intelligence (AGI) – the idea of computers that can understand the world like humans do.

“Vision is one of our most important senses,” Zeng says. “If we want AI to truly think and learn like us, it must be able to see like us – or even better.”

Why This Is Good News

Thanks to this work, imaging systems can become smarter, faster, and more flexible – whether it’s in a hospital, a field, or your pocket. These improvements help us monitor health, protect the environment, and even understand the world in entirely new ways.

This research also supports the UN Sustainable Development Goal 9 (Industry, Innovation and Infrastructure) by making cutting-edge technology more reliable, more efficient, and more useful in real-world situations.

Read a more detailed summary or the entire PhD

-

PhD Title: Multispectral Image Reconstruction and Restoration in Computational Imaging with Applications in Photography

-

Promotors: Wilfried Philips and Hiep Luong