Research paves the way toward real-time CGI
If a picture paints a thousand words, then two pictures paints – a computer generated hologram?
In the drive to make three-dimensional imaging and graphics more efficient, and someday even instantaneous, researchers from The Ohio State University are pushing the boundaries.
The research of Sihao Ding, a PhD student in electrical and computer engineering at Ohio State, was recently published in the Optical Society’s Applied Optics journal, titled “From image pair to a computer generated hologram from a real-world scene.”
Essentially, Ding said, his research explores the creation of 3D computer generated holograms utilizing just two photographs from a camera. The process helps do away with expensive laser setups, simplifying it down to a couple photographs.
“If you want a hologram of a real world scene you don’t have to set up all the physical equipment with laser and all that,” Ding said. “You just use cameras, take two images and you can get the hologram.”
While the concept of creating holograms from photography is not new, Ding said, his research helps advance the efficiency and mathematics behind it.
“This is the first work,” he said, “that goes directly from two images to the computer generated hologram.”
To model a 3D object, Ding said, two cameras are first set up to take pictures of the same scene from mirror sides. The images are then uploaded into a computer and different parts of the scene are selected in what is called a “point cloud,” used to generate the image in computer form. He said each point in the cloud represents a light source that emits waves, propagating toward a receiving plane.
“We regenerate them and reconstruct them from different views and focal distances,” he said.
Positioning the initial camera shots is important in the process.
“There needs to be some kind of geometry set up. It’s actually called a stereo pair,” he said. “From these images, we can calculate the depths of the 3D scene. This is basically the first step and after we get this we can compute a 3D model of the whole scene.”
The research was written by Ding, with the assistance of his advising ECE professor Yuan Zheng, as well as scholars from the University of Arizona and Air Force Research Laboratory.
Ultimately, the research helps improve upon the algorithms used to create the point cloud images.
Creating more efficient holograms, Ding said, is helpful in the further development of the augmented or virtual reality displays of the future.
Not only does the process streamline the process, it saves money.
Overall, he said, the research behind making more efficient holograms remains in its infancy. While his work pushes the boundaries further out, the quality and detail of the hologram created by using two images must improve.
“We still have a lot to do to improve the quality, and still this cannot achieve real time,” Ding said. “But the potential is there because the algorithm is set up so it’s possible to parallel the computing.”
His research, however, can help bridge the gap between work being conducted by others in his field.
“Maybe one day it can be done in real time,” Ding said
