Want to see an old family film in 3D, or the holiday pictures you took with your smartphone? That might be possible soon. Delft researchers developed a method to give images depth, with a little help from 17th-century painters.
The ancient Greeks already realised that in order to paint depth, one must use the fact that everything further away is darker and blurrier. Medieval artists forgot this. Hence, medieval paintings appear flat. But in the Renaissance the school of Caravaggio rediscovered this "claroscuro" or "clair-obscur". The Flemish and Dutch painters of the 17th century followed suit.
Now Delft researchers too use this technique. Professor of vision based robotics, Pieter Jonker (3mE faculty), his assistant Boris Lenseigne and his students Eelko van Breda and Vikko Smit, developed a method using luminescence to devise a depth map. From the depth map, one can automatically create various other 3D formats to watch your content on your 3D TV, your VR headset, with old-school red-blue goggles or with a cardboard or plastic stereo viewer in front of your smartphone.
"In modern imaging, depth from luminance is not being used", said Jonker. "We are so focused on the idea that we need two eyes to see depth." And that is a shame, Jonker believes. Two dimensional pictures contain the information of the original 3D world they depict. That information, relative depth, is encrypted in the amount of luminance spread out in the image.
An important aspect of the Delft technology is a way to filter out lighting artefacts in images. "Our eyes evolutionary developed with only one light source, the sun", said Jonker. "In that situation the fact that everything further away is darker holds, only shading and shadows ruin this law. We developed a method to filter out reflections and shades from multiple light sources, resulting in an image that looks like London in the fog. It is grey, there are no shiny surfaces, and the light comes from nowhere and everywhere. In such situations the depth-from-luminance or the clair-obscur method holds."
The technology is patented and the just started TU Delft spin-off company QdepQ will allow people to upload images on their website to turn them into 3D format or upload them via an Andriod app later this year. An app for iOS is in development.
Returning to the painters of the 17th century, Rembrandt was a master in clair-obscur. He deliberately painted the two people closest to the front in the scene in Nachtwacht entirely white and wearing completely black robes. "He also put multiple light sources in the painting, notably on a girl in a white dress. He probably wanted to show that darker images which are far away are not a pure cue for depth. Our brains have learned to cope with these exceptions. We use contextual information as well," said Jonker.
Jonker tested his method on an image of the Nachtwacht and was amazed to see how well Rembrandt could paint depth. "And", he said, "how well the patented method works."