Joint Decision Tree and Visual Feature Rate Control Optimization for VVC UHD Coding.

Area of Science:

• Video Compression
• Image Processing
• Computer Vision

Background:

• Ultra-high-definition (UHD) video coding demands efficient rate control to manage large data volumes.
• Existing methods for versatile video coding (VVC) may not fully leverage visual features for optimal rate-distortion performance.

Purpose of the Study:

• To develop an advanced rate control scheme for UHD VVC by integrating decision trees and visual feature optimization.
• To enhance the accuracy of rate-distortion (R-D) models for UHD video compression.

Main Methods:

• A novel R-D model tailored for UHD videos was designed.
• A decision-tree-based multiclass classification scheme was established, incorporating visual features to improve R-D model prediction accuracy.
• Convex optimization was employed to find the globally optimal solution based on the proposed R-D model.
• The algorithm was integrated into the VVC Test Model (VTM) 10.2 reference software.

Main Results:

• The proposed method achieved significant bit rate reductions compared to the latest VTM 10.2 algorithm and other state-of-the-art approaches.
• Quality metrics, including Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM), were maintained at given levels.
• Experimental results demonstrate the effectiveness of the joint decision tree and visual feature optimization approach.

Conclusions:

• The developed rate control scheme offers substantial improvements in bit rate efficiency for UHD VVC.
• The integration of decision trees and visual features provides a robust method for optimizing video compression.
• This approach represents a significant advancement in UHD video coding technology.