Intra Predictive Depth Map Coding Using Flexible Block Partitioning.

Area of Science:

• Computer Vision
• Image and Video Compression
• Signal Processing

Background:

• Depth map compression is crucial for 3D video applications.
• Existing methods like 3D-HEVC have limitations in intra-coding efficiency.
• Depth maps exhibit unique characteristics, such as smooth regions and sharp edges, requiring specialized compression.

Purpose of the Study:

• To propose an efficient intra-based depth map compression algorithm.
• To develop a novel predictive depth coding (PDC) method.
• To improve rate-distortion performance for depth map coding.

Main Methods:

• Developed a directional intra prediction framework with optimized block partitioning.
• Introduced a constrained depth modeling mode for edge representation.
• Utilized constant and linear residue modeling for low-complexity residue coding.

Main Results:

• PDC demonstrated superior rate-distortion performance compared to 3D-HEVC intra coding.
• Achieved average bitrate savings of 6% without view synthesis optimization.
• Achieved average bitrate savings of 14.3% with view synthesis optimization.

Conclusions:

• PDC is an efficient alternative for intra depth map coding.
• The method effectively handles smooth areas and sharp edges in depth maps.
• The proposed approach offers significant bitrate savings without transform-based residue coding.