A Fast Decision Algorithm for VVC Intra-Coding Based on Texture Feature and Machine Learning

Area of Science:

• Digital video compression
• Information technology
• Machine learning applications

Background:

• Modern information technologies like 5G, big data, and AI increase demands on video coding.
• High-Efficiency Video Coding (HEVC) has limitations in block division flexibility and prediction detail.
• Versatile Video Coding (VVC) offers improved coding efficiency but introduces high computational complexity due to its advanced block division structure (QT and MTT).

Purpose of the Study:

• To reduce the computational complexity of Versatile Video Coding (VVC) intra-frame coding.
• To develop a fast decision algorithm for VVC coding block division.
• To enhance coding efficiency without significant bit rate penalty.

Main Methods:

• Proposed a fast decision algorithm for VVC intra-frame coding.
• Analyzed Characteristics of CU partition structure decisions.
• Utilized texture complexity for early termination of the CU partition process.
• Employed a three-category feature-trained tandem classifier using global sample, local sample, and context information for CU division type prediction.

Main Results:

• The proposed algorithm achieves a 1.36% increase in encoding output bit rate compared to VTM10.0.
• Significant saving of 52.63% in encoding time was observed in full intra mode.
• The algorithm effectively reduces computational complexity in VVC intra-frame coding.

Conclusions:

• The developed fast decision algorithm successfully reduces VVC coding complexity.
• Texture characteristics and machine learning provide an effective approach for optimizing VVC block partitioning.
• The proposed method balances coding efficiency and computational cost for advanced video coding standards.