US University and IIT Madras Develop AI Powered Algorithm to Amplify 3D Effect in Smartphone Videos

Deep learning algorithms developed by researchers at IIT Madras and Northwestern University in the United States can significantly increase depth perception and 3D effects in videos taken with smartphone cameras.

Checkout: Northwestern University Programs

As per the officials, such algorithms will prevent mobile phone photos from becoming "flat" and will give them a real 3D feel. One of the biggest advantages of the algorithm is that it rectifies the need for an expensive equipment or hundreds of lenses to shot videos with depth.

According to Kaushik Mitra, assistant professor at IIT Madras' Department of Electrical Engineering, "Photographs and movies shot on smartphone cameras have a flat, two-dimensional aspect, which is a common problem among professional photographers. Apart from the flat appearance, several 3D characteristics, such as the 'Bokeh Effect,' which is easier with a DSLR camera but difficult with a smartphone camera, are hard to obtain". 

Check: Northwestern University Admission 2022

"While a few mid and high-end smartphone cameras have now been programmed to include similar effects in stationary images, particularly in portrait mode," he continued, "it is not yet possible to reproduce them in movies shot using smartphones."

"LF capture is achieved by using a series of microlenses placed between the camera's primary lens and the sensor. Due to space limits, many micro lenses cannot be utilized on mobile phones. Instead, techniques for post-processing photographs taken with existing smartphone cameras are being developed.

Read:  US Most Likely to Extend International STEM Graduates’ Employment Visa Duration

"For picture alteration, artificial intelligence and machine learning techniques are applied. Our team investigated the problem and developed a deep learning algorithm that converts stereo photographs taken with a smartphone into LF images "Mitra remarked.

"The research was published in the proceedings of the International Conference on Computer Vision (ICCV), 2021," he continued." The algorithm begins by simultaneously capturing two movies (known as a stereo pair) with the two neighboring cameras found in many modern smartphones. Deep learning models are used to take these stereo pairs through a series of phases. The stereo pairs are transformed into a 7X7 grid of pictures, simulating a 7X7 array of cameras, and the LF image is created ".

"A key benefit of our algorithm is that it eliminates the need for expensive equipment or a variety of lenses to create videos with depth." A smartphone with a dual-camera system may generate Bokeh and other attractive 3D effects. "In addition to giving depth, our algorithm allows us to see the same video from any of the 7x7 grid of viewpoints," he noted.

Also Read:

• In-person Visa appointments at US Embassy in India closed for now
• ASU Aims to Enroll 100 Million Students by 2030