Master Thesis: Optimizing Temporal Noise Reduction on a smartphone using Halide at IMINT

Real-time on-device Temporal Noise Reduction has become a common feature on modern smartphones. Processing video frames with high resolutions and frame rates is an expensive process however, and the power consumption requirements are extremely strict. The goal of this thesis is to investigate the possibilities of porting an OpenGL-based Temporal Noise Reduction solution to the Halide language, which can be accelerated by a Hexagon DSP on Qualcomm chipsets. This includes porting of the OpenGL source code and analyzing the performance.

Attention: Often you need a pre-approval from your university or study counselor, to ensure that projects or thesis found on AAU Jobbank will be accepted as part of your education. Please contact the right entity in due time to ensure that you're picking the right project.