Do receives grant to advance 3-D image, video processing for consumer devices
Elise King, CSL Communications
- ECE Associate Professor Minh Do recently received a grant from the National Science Foundation for research on image and video processing with 3-D cameras.
- Do believes this technology could be successful if implemented in more consumer products.
- Do's research will focus on improving the quality and resolution of images, and on integrating multiple depth frames.
ECE Associate Professor Minh N. Do recently received a 3-year, $390,510 grant from the National Science Foundation to advance the theory and practice of image and video processing with 3-D cameras.
Nowadays smartphones and laptops are all equipped with built-in cameras; however, these are just color cameras. Do’s research will focus on advancing color and depth cameras, such as ones used in the popular Microsoft Kinect technology, with the ultimate goal of making these cameras more widely available for consumer use.
Do, a researcher in the Coordinated Science Lab (CSL), predicts this type of technology will be very successful if used in more consumer products. Kinect is helping pave the way.
“What we have now with Kinect is a camera that is very cheap," Do said. "It became the fastest-selling consumer electronic device, selling 2.5 million units in the first 25 days after introducing it to the market.”
These types of cameras, which combine depth and color information in real time, can be used to capture live 3D images. For example, they could be used to create a 3-D model of a person and then insert them into a virtual reality.
However, in order to use these cameras in more everyday devices, researchers must develop fundamental new methods for processing depth input in combination with regular color images and videos. Essentially, Do will focus on how to advance the signal, image and video processing theory and algorithms to deal with these types of new data.
“One fundamental problem with these devices, and like other type of data that you capture live, is they have a lot of physical imperfections," Do said. "For example, the signal you measure is noisy. So one of the key challenges is to enhance and clean up the data being captured."
The specific research will involve four different components. First, researchers will look at calibrating and denoising raw measurements from depth cameras using accurate physical and statistical models. They will also focus on improving the quality and resolution of these images, and they will integrate multiple depth frames to extend super-resolution to video. Finally, they will use this improved depth information in challenging and high-impact applications.
Do is the principal investigator on this project and Sevket Babacan, a former postdoctoral fellow at the Beckman Institute for Advanced Science and Technology, will be the co-PI. Do’s research area at CSL is in signal and image processing.