Enhanced Cross-Component Linear Model for Chroma Intra-Prediction in Video Coding.

Area of Science:

• Video Compression Technology
• Digital Signal Processing
• Computer Vision

Background:

• The Cross-Component Linear Model (CCLM) is a key tool in the Joint Exploration Model (JEM) for chroma intra-prediction.
• CCLM leverages linear correlations between luma and chroma components for efficient prediction.
• Existing CCLM methods can be further optimized for enhanced coding efficiency.

Purpose of the Study:

• To introduce novel methods for improving the coding efficiency of CCLM in video compression.
• To enhance the prediction accuracy of chroma components using advanced linear modeling techniques.
• To contribute to the development of more efficient video coding standards.

Main Methods:

• Developed a multi-model CCLM (MM-CCLM) approach, classifying neighboring samples into groups for tailored linear model parameter derivation.
• Introduced a multi-filter CCLM (MF-CCLM) technique, enabling optimal down-sampling filter selection for the luma component in 4:2:0 format.
• Proposed a LM-angular prediction (LAP) method, integrating angular intra-prediction with MM-CCLM for a novel chroma intra coding mode.

Main Results:

• Achieved average BD rate savings of 0.55% for Y, 4.66% for Cb, and 5.08% for Cr components under All Intra (AI) configurations.
• Demonstrated significant bitrate reduction through the proposed MM-CCLM, MF-CCLM, and LAP methods.
• Validated the effectiveness of MM-CCLM and MF-CCLM, leading to their adoption into the JEM by JVET.

Conclusions:

• The proposed MM-CCLM, MF-CCLM, and LAP methods offer substantial improvements in chroma intra-prediction efficiency.
• These advancements contribute to reduced bitrate requirements for video compression.
• The integration of these techniques into JEM highlights their practical significance and effectiveness in state-of-the-art video coding.