Efficient Intra Mode Selection for Depth-Map Coding Utilizing Spatiotemporal, Inter-Component and Inter-View...

Area of Science:

• Computer Science
• Electrical Engineering
• Image Processing

Background:

• 3D-high efficiency video coding (HEVC) is crucial for compressing multi-view video plus depth (MVD) data.
• Current 3D-HEVC intra prediction modes, including depth-modeling modes (DMMs), drastically increase computational complexity.
• The depth intra mode decision process is time-consuming due to complex rate distortion (RD) cost calculations.

Purpose of the Study:

• To propose a low-complexity intra mode selection algorithm for 3D-HEVC.
• To reduce the computational complexity of depth intra prediction in both intra- and inter-frames.
• To maintain high video quality while optimizing coding efficiency.

Main Methods:

• Analyzed inter-view and inter-component correlations in intra coding information (intra mode, RD cost).
• Classified intra modes into three activity classes with distinct mode-weight factors.
• Defined coding mode complexity based on neighboring coded CUs and utilized it to assign candidate intra modes.
• Developed a method to skip unnecessary intra prediction sizes using optimal mode and RD cost.

Main Results:

• Achieved a 61% reduction in intra prediction complexity.
• Incurred only a 0.2% Bjontegaard metric increase for coded and synthesized views.
• Demonstrated the effectiveness of the proposed fast depth intra coding algorithm.

Conclusions:

• The proposed algorithm effectively reduces computational complexity in 3D-HEVC depth prediction.
• The approach balances significant complexity reduction with a negligible impact on video quality.
• This method offers a practical solution for efficient 3D-HEVC encoding.