A Game Theory Based CTU-Level Bit Allocation Scheme for HEVC Region of Interest Coding

Area of Science:

• Computer Vision
• Video Coding
• Game Theory

Background:

• Subjective video quality is crucial for user experience in video conferencing.
• Efficient bit allocation is essential for optimizing video compression performance.
• Existing methods may not adequately address quality disparities between important and unimportant video regions.

Purpose of the Study:

• To develop a CTU-level bit allocation scheme for subjectively optimized video coding.
• To improve the visual quality of Regions of Interest (ROI) in video conferencing.
• To establish a framework for balancing ROI quality with overall video quality.

Main Methods:

• Formulation of a non-cooperative Stackelberg game for bit allocation.
• Defining video regions as game players (ROI as leader, non-ROI as follower).
• Solving a utility optimization problem to determine bit allocation strategy.
• Application of a Newton-method-based algorithm for CTU-level bit distribution.

Main Results:

• The proposed scheme effectively allocates bits between ROI and non-ROI regions.
• Significant improvements in ROI visual quality were observed.
• Acceptable overall video quality degradation was achieved.
• Experimental results validated through objective and subjective assessments.

Conclusions:

• The Stackelberg game-based bit allocation scheme enhances subjective video quality in conferencing.
• Prioritizing ROI leads to a superior visual experience for users.
• The method offers an effective trade-off between specific region quality and global video fidelity.