Real-time FPGA-based Architecture for Image andVideo Upscaling from Local Self-Examples

This project is about image upscaling which refers to the task of constructing a high
resolution image of a given lower resolution image, while preserving features of the original
image, such as sharpness and texture. One of the utmost difficulties that limit practical
application of state-of4he-art upscaling algorithms is their inherent computation demand, and
thus, conventional software implementations of upscaling algorithms are often unable to
deliver the required performance in many commercial applications such as mobile devices,
digital TVs, media players, and satellite imaging. In this project, we aim at a commercially
viable hardware (FPGA-based) implementation of an upscaling method that exploits local
self-similarities in the input image. We propose to desig a resources and power optimized
FPGA-based reconfigurable IP core that shall upscale HD1 D8Dp video streams in real-time
(30 frames/s). Our proposal outlines how we plan to achieve such a commerciallP core for a
robust image upscaling.

Faculty Supervisor:

Aishy Amer

Student:

Partner:

TandemLaunch Inc

Discipline:

Engineering

Sector:

Finance and Insurance; Professional, scientific and technical services

University:

Concordia University

Program: