US University, IIT Madras Researchers Develop AI Technology for Enhanced Videos on Smartphones


Sristy Sharma
Updated on Mar 21, 2022 02:42 PM IST

Researchers at US' Northwestern University and IIT Madras collaborated to build deep learning algorithms that will prevent mobile images from appearing "flat" and give a realistic 3D feel.

US University, IIT Madras Researchers Develop AI Technology for Enhanced Videos on Smartphones

A collaboration of researchers at IIT Madras and Northwestern University (US) has resulted in the development of AI-based deep learning algorithms, which can greatly enhance the depth perception and 3D effects in videos shot from smartphone cameras.

Kaushik Mitra, Assistant Professor (Department of Electrical Engineering, IIT)stated that the technology can help in upgrading the flat impression of current smartphone images and ingivinga realistic 3D impression.

The complaint is common, especially among amateur and professional photographers, that smartphone captures give out a flat and 2D-look to the objects. The existing 3D features, such as the Bokeh Effect (aesthetic blurring of background), are easierto manage with DSLR cameras. It was acknowledged that while certain effects are available in portrait mode onhigh-end smartphones,video captures on smartphones are still away from the technology.

Also Read:Truman State University Unveils Scholarships for Meritorious Indian Students

According to Mr Mitra, professional cameras use information about the intensity and direction of light (Light Field or LF) to give the perception of depth. In order to capture LF, an array of microlenses are inserted between the main lenses and camera sensors, which is a challenge in mobile phones due to space constraints.

AI and machine learning are being used to develop algorithms that can post-process the captures by mobile phones. The technology can help in better image manipulation, which has been used to develop an algorithm that changes stereo images into LF images.

The algorithm simultaneously captures two videos (stereo pair) from adjacent cameras present in smartphones, and processes them through a sequence of steps, based on deep learning models. The stereo pair gets converted into 7X7 grid images, mimicking a 7X7 array of cameras, and produces LF images.

Additionally, processed video can be viewed from any point of view available in the 7X7 grid.

Also Read:University of Manchester Announces STEM Scholarships for Indian Women

The research has been published in Proceeding of International Conference on Computer Vision (ICCV) 2021.

The technology shall enable all dual-camera equipped phones to capture better depth perception images and ease the use of Bokeh and other 3D aesthetic effects usage. This technology will eliminate the need for fancy and expensive equipment, or lenses to capture videos with depth.

Get the Best Assistance in Applying toTop US Universities!

Source: NDTV Education

You can share this post!

Admission Open 2024

Related Articles

Aug 18, 2023 04:28 PM IST

How to Apply for Masters in USA?

Aug 03, 2023 01:18 PM IST

Documents Required for TOEFL

Our Study Abroad Offerings

Related Institutes

Planning to study abroad?

Interested in Studying Overseas?

Connect with us and kickstart your learning journey for a rewarding experience!

By proceeding ahead you expressly agree to the CollegeDekhoAbroad Terms of Use & Privacy Policy.