Date of Submission
Fall 8-16-2018
Degree Type
Thesis
Degree Name
Master of Science in Computer Science (MSCS)
Department
Computer Science
Committee Chair/First Advisor
Prof. Ming Yang, Prof. Ying Xie
Track
Big Data
Chair
Prof. Ming Yang
Committee Member
Prof. Ming Yang
Committee Member
Prof. Ying Xie
Committee Member
Prof. Selena He
Abstract
High efficiency Video Coding (HEVC) has been deemed as the newest video coding standard of the ITU-T Video Coding Experts Group and the ISO/IEC Moving Picture Experts Group. The reference software (i.e., HM) have included the implementations of the guidelines in appliance with the new standard. The software includes both encoder and decoder functionality.
Machine learning (ML) works with data and processes it to discover patterns that can be later used to analyze new trends. ML can play a key role in a wide range of critical applications, such as data mining, natural language processing, image recognition, and expert systems.
In this research project, in compliance with H.265 standard, we are focused on improvement of the performance of encode/decode by optimizing the partition of prediction block in coding unit with the help of supervised machine learning. We used Keras library as the main tool to implement the experiments. Key parameters were tuned for the model in our convolution neuron network. The coding tree unit mode decision time produced in the model was compared with that produced in HM software, and it was proved to have improved significantly. The intra-picture prediction mode decision was also investigated with modified model and yielded satisfactory results.