Have you ever taken a picture, only to find it turned out blurry?
Maybe the camera shook, or the person you were photographing moved at just the wrong moment.
Engineers at Johns Hopkins University have come up with a smart solution that can make these fuzzy photos sharp and clear, and they’re calling it the Progressively Deblurring Radiance Field, or PDRF for short.
This new tool is not only super fast, making images clear 15 times quicker than older methods, but it also works wonders on both pictures that are made-up (synthetic) and those taken in the real world.
Cheng Peng, a clever researcher at Johns Hopkins, explained that this technology could turn blurry images into sharp, three-dimensional ones.
This could be a big deal for all sorts of things – from virtual reality games to online shopping, making movies, helping robots find their way, and even just improving your own photos and videos.
Peng and his mentor, Rama Chellappa, shared their findings at a big conference on artificial intelligence, showing how well their method works.
Usually, making a blurry image clear is a two-step dance. First, the computer has to guess where the camera was when it took the blurry picture. Then, it uses that guess to build a 3D scene from the 2D photo.
While this approach sort of works, it’s not perfect and can leave weird glitches or miss out on some details.
There’s been a recent leap forward with something called Neural Radiance Field (NeRF) that makes super realistic 3D images, but it needs high-quality photos to start with.
PDRF, however, can take even not-so-great photos and make them sharp before turning them into 3D scenes.
It’s like having a smart assistant that not only fixes the blurriness but also cleans up the image, all thanks to some very clever computer brainpower. This method learns by itself from the images you give it, meaning it doesn’t need any extra help or examples to get going.
For instance, this new tech is being used to help doctors better see and understand skin tumors, which can be really tricky with traditional methods.
Peng and his team are working on making detailed 3D models of these tumors, which could be a big help in diagnosing and treating them, especially through telemedicine.
This groundbreaking work has even caught the attention of some big names in research, earning support for projects that could one day let us explore detailed 3D models of cities and far-off places, all from our own homes.
Imagine being able to travel to distant lands or walk through historical cities in 3D, all from pictures taken by people just like you. Thanks to the engineers at Johns Hopkins, that future might be closer than we think.
Source: Johns Hopkins University.
