Adaptive Progressive Motion Vector Resolution Selection Based on Rate-Distortion Optimization

Area of Science:

• Video Compression
• Digital Signal Processing
• Computer Vision

Background:

• The H.265/HEVC standard uses a fixed 1/4-pel motion vector (MV) resolution, potentially limiting motion compensation accuracy.
• This fixed resolution overlooks variations in video content like texture complexity and motion activity.
• Existing methods may not meet the demands for high-accuracy motion compensation in diverse video sequences.

Purpose of the Study:

• To develop an adaptive motion vector resolution selection scheme for H.265/HEVC video coding.
• To enhance coding performance by considering inherent video characteristics.
• To improve the accuracy of motion compensation through content-adaptive MV resolution.

Main Methods:

• A novel rate-distortion model was designed to evaluate MV resolution candidates based on video characteristics.
• Motion vector resolution selection was framed as a rate-distortion optimization problem.
• A progressive MV resolution strategy, guided by decision trees derived from the rate-distortion model, was employed.

Main Results:

• The proposed algorithm adaptively adjusts MV resolution based on local content properties.
• Experiments demonstrated significant improvements in coding performance.
• An average bitrate reduction of 1.8% (BD-rate gain) was achieved.

Conclusions:

• The developed adaptive progressive motion vector resolution selection scheme effectively enhances H.265/HEVC coding efficiency.
• The method achieves notable performance gains without increasing computational complexity.
• Content-adaptive MV resolution selection is a viable strategy for improving video compression.