Optimizing encoder configuration for real time video transcoding

An advanced video encoder (e.g., HEVC, AVC), has many encoding configuration parameters. Encoding “presets” set the values of certain codec parameters and thus facilitated configuring the encoder. The goal of this project is to develop a computationally efficient learning-based approach to make a run-time decision on the encoder’s optimal preset configuration to achieve the best quality for a given bit rate. Since different presets have different transcoding time for different videos, we will develop a machine learning method to estimate the transcoding time for a video segment based on its content for different presets. We will also predict the Rate-Distortion (R-D) characteristics of the video segment for each preset. This way, without transcoding, we will have quality-bit rate-computation time information about a video segment. We will use the quality-bit rate-computation time information to maximize the quality of the transcoded video segment while maintaining a real time performance.

Faculty Supervisor:

Shahram Shirani

Student:

Partner:

Amazon Canada

Discipline:

Engineering

Sector:

Information and cultural industries

University:

McMaster University

Program: