Efficient Depth Intra Frame Coding in 3D-HEVC by Corner Points.

Area of Science:

• Computer Vision
• Video Compression
• 3D Graphics

Background:

• 3D-HEVC depth coding faces complexity challenges due to flexible partitioning and numerous intra modes.
• Existing methods struggle with multi-directional edges common in depth maps.
• Depth maps exhibit unique characteristics like large plain regions and sharp object boundaries.

Purpose of the Study:

• To reduce the computational complexity of 3D-HEVC depth intra coding.
• To improve the prediction accuracy for Coding Units (CUs) with complex edges.
• To develop an adaptive algorithm balancing quality and complexity.

Main Methods:

• Utilizing corner points (CPs) to guide the splitting of CUs with multi-directional edges.
• Implementing a fast intra mode decision for non-CP Prediction Units (PUs).
• Developing a two-step adaptive CP selection technique for content and quantization parameter adaptability.

Main Results:

• Achieved approximately 66% time reduction in the 3D-HEVC intra encoder.
• Demonstrated no noticeable performance degradation in synthesized view quality.
• Outperformed prior state-of-the-art algorithms in time reduction and delta Bit Rate (∆ BDBR).

Conclusions:

• Corner points are effective features for optimizing depth map coding.
• The proposed method offers a significant speed-up for 3D-HEVC intra encoding.
• The algorithm provides a flexible trade-off between synthesized view quality and computational complexity.