Neural implicit functions for multi-image low-level computer vision on smartphones

Modern smartphone cameras commonly employ multi-image (or burst photography) for tasks related to image super-resolution and high-dynamic-range imaging. This project is focused on developing novel multi-image techniques that leverage the power of recently proposed neural implicit functions. Neural implicit functions (NIFs) are a new way to represent images not as a 2D grid of pixels values but instead as functions. This functional representation offers many benefits, including a continuous representation, effective interpolation, and compact representation. We are interested in using NIFs for burst photography. The goal is to fuse the multiple images into a new image with enhanced properties, such as improved image detail (super-resolution) or higher dynamic range (HDR). This project will help Samsung Electronics Canada develop improved camera performance for its smartphone devices.

Faculty Supervisor:

Michael S. Brown

Student:

Partner:

Samsung Electronics Canada

Discipline:

Computer science

Sector:

Manufacturing

University:

York University

Program: